I mentioned in a previous post that my “work life had shifted recently.” That’s actually the polished, PR version of the sentence. The real version is: I got laid off. 😒

The company I’ve been working for recently laid off its entire team to restructure under new ownership. And while I know it’s just business, it’s still disorienting. But instead of spiraling, I’m choosing to look forward—focusing less on "job security" and more on "career durability."

For this next phase, that means staying sharp, staying connected, and staying on top of where our industry is going. Coincidentally, that’s also the exact reason this newsletter exists.

Juggling this publication while actively searching for my next role has required a bit more discipline than usual, but in many ways, the timing feels clarifying. If I’m going to write every week about how AI is reshaping communications, marketing, media, and hiring, I can't think of a better moment to put those exact ideas to work on my own career.

So, let’s get into this week’s issue!

💻 I Built a Site With ChatGPT, and I Didn't Write a Line of Code

ChatGPT’s Sites is OpenAI’s new website and web app builder inside ChatGPT Work that went live recently in public beta via paid plans.

Basically, you describe what you want in your website in plain English, and ChatGPT builds it, hosts it, and gives you a live link. No code, no separate hosting account, no calling in a favor from your friend who “knows WordPress.”

🧪 Beta Lab: Building a Site (using ChatGPT “Sites”)

Goal: Build a professional portfolio site to support an active job search, without hiring a designer or learning to code.

Tools: ChatGPT Sites (ChatGPT Work via ChatGPT Plus plan or Codex)

The Experiment:

• First, open ChatGPT and toggle into “Work” (or, in my case because Work hasn’t rolled out to my ChatGPTyet, I am using Codex for this).

• Make sure any plugins you’ll need are turned on. Since I wanted to build a professional website and portfolio of my work, I needed it to have access to my Linkedin, Substack, and Google Docs.

• Second, if Sites has rolled out to your ChatGPT work and it’s showing in the menu, great. But, you can also type “/” and look for “Sites” in the command menu, or just ask ChatGPT to build you a “site” and (hopefully, if it’s rolled out to your account), it will trigger automatically.

• Third, typed in your prompt. For this, mine was:

  “Please create a @Site. I’m interviewing for new roles, and I want to create a page/portfolio of my work. I want the site to contain an executive bio, three leadership case studies from different points of my career with a clear problem/decision/outcome structure, examples of thought-leadership work and my communications philosophy, links to several of my strongest newsletter pieces, and two or three concise testimonials.”

The Results: It took me a full(ish) workday (minus several work calls/schedule interruptsions) of going back and forth, the same way you’d iterate with a human web designer, but the difference is that I didn’t need my content pre-written or packaged in an way. I simply pointed it toward my LinkedIn, this Substack, and a folder of documents, and it pulled, drafted, and organized it all from there.

🌟 SO …. drum roll, please, here is my site, and I’m proud of it!

The Verdict: Totally worth it. It’s like magic … you must try it.

A few things worth knowing before you try it:

• Iterations cost credits every single time you make a change. If you’re on a Plus plan working solo like I am, you’ll either need to try to stay within your allotted usage limits or buy credits, and credit packages start at $100 for 2,500 tokens (!). Also, you’ll need a ChatGPT Business plan if you want to collaborate with someone else inside of the site itself.

  • There are many great tips here on how to iterate more efficiently and using less credits. I found this Mindstudio article to be very helpful in general with tips about this.

• It defaults to its own design opinion. If you don’t specify colors or style, it picks something. The genius part is you can just ask it to change anything (colors, icons, layout) and it does it on the spot. Each change costs tokens, though, so don’t get precious about tweaking forever. And, if you have specific branding or color codes you want it to use, you’ll want to include that information in your original prompt.

• Updates are live. Once published, any change you request updates the site in real time. You never have to hit “publish” again!

Once you’re ready to go, you get a hosted URL by default, or you can connect a custom domain through your registrar’s DNS settings. You can also choose your access level, whether you want it fully private, workspace-only, invite-only, or public. OpenAI then handles hosting natively, but if you want to self-host elsewhere, you can request the raw HTML, CSS, and JavaScript and move it to Wix, Squarespace, or whatever platform you’re using.

It is worth noting that OpenAI’s terms block Sites from processing payment-card data or protected health information, so you can’t use it for financial transactions.

💡 Beyond a job-search portfolio, a few other practical uses for Sites include:

• A live dashboard to track how an announcement is performing that you can share with others within your workspace.

• A microsite for event attendees to get all of the information they need in one, visually-appealing location instead of a static PDF.

• A reporter/executive brief for a interview preparation leading up to media interviews or a big announcement.

📌 If you want help building your own site, I’ve begun coaching individuals and teams on this and AI communications workflows. Let’s talk.

🔨 The AI Slop Crackdown Has Officially Arrived, and I Have Feelings About It

LinkedIn just rolled out a “Seems like AI slop” flag that users can attach to posts and comments they think are low-effort, AI-generated content. Substack, meanwhile, partnered with Pangram, an AI-detection company, to start identifying AI-generated posts on the platform.

On the surface, I get it. Both platforms are drowning in generic AI output, and both companies (and the people on these platforms) have a business reason to want their content to feel human again.

But, here’s my problem: I personally love to write in short sentences. I frequently use em dashes. I love to use fragments for emphasis and comparisons and series, and basically everything that is oftentimes an AI slop “tell.” I’ve written this way for twenty-plus years (I mean, I majored in Journalism from UGA, go dawgs!), long before ChatGPT existed, and now that exact style is getting flagged, along with words such as “quietly,” “amid,” and “delve.” So, apparently my natural voice and a language model’s average output have converged, which is either flattering or deeply annoying.

I watched this YouTube video by Dr. Taylor Jones (@languagejones), a linguist, breaking down how to detect AI slop, and he offered some interesting context for me on this topic. He suggested that beyond the “tells” that we’re all aware of (above), the worst and most glaring violation of AI-enhanced writing is the total incoherence hiding behind confident, fluent word choices and the sentences generated that are completely nonsensical. And, I wholeheartedly agree that this should be considered a far more egregious/flagged “tell” of AI slop than the overuse of an em dash.

All this to say that I am genuinely worried that this will all turn into a witch hunt, where distinctive human writers get flagged right alongside the actual slop, and everyone starts sanding down their voice to avoid the accusation.

For what it’s worth, here’s my actual disclosure: yes, I use AI when I write this newsletter. I mean, I’m writing about AI … it would be a little absurd not to. My process is that I write, from scratch (!), an outline and rough draft of the piece. Then, I run it through Claude and specifically ask it to “take this first draft of my next substack post and polish it up. Please also flag anything that doesn’t make sense or is gramatically incorrect.” Then, I use Gemini to help me draft the SEO headline and content tags.

In other words: Me/written draft > Claude > Gemini > Substack

So, there you have it. That’s my workflow, and I’m not hiding it.

🔄 3 AI Workflows I’m Actually Running Right Now

I’m often asked what “using AI at work” actually looks like in my day to day. So, to answer that question, here are three workflows I’m currently running that you may find helpful, too.

Brand Partnership Analytics

I have several consulting projects going and one of them is that I’m working with a social media creator on brand partnerships. So, I’ve created a shared Google folder for her to upload screenshots of her social platform analytics. Then, I prompt ChatGPT Work to review the files in that folder and translate the raw numbers into a practical positioning brief including the audience profile, engagement quality, and strongest selling points. I've then automated it to upload that output directly into a Notion project workspace.

Flow: Google Drive screenshots → ChatGPT Work → Notion brand folder

Building a Thought Leader’s Platform

I’m also consulting with a writer who is writing a book and publishing proposal. We needed a system for our weekly calls, notes, and task tracking that didn’t rely on memory or a scattered inbox. Now our weekly Google Meet calls get transcribed by Granola, which I’ve automated to feed the notes into our shared Notion workspace, and tasks from the call get assigned specific owners and important due dates.

Flow: ChatGPT Work → Google Meet → Granola → Notion

Job Opportunity Digest

Given the job search I mentioned earlier, I needed one place to see everything instead of checking LinkedIn, Indeed, and several professional Slack channels I’m a member of separately. So, I set up ChatGPT Work to pull all of the listings from all three sources into a single 8am digest that lands in my Gmail inbox daily.

Flow: ChatGPT Work → LinkedIn, Indeed, Slack → Gmail inbox

The Bottom Line

If there’s a theme running through everything I’ve been doing lately, it’s experimentation.

I built a website because I needed one. I’ve been testing AI workflows because I want to understand which ones actually save time (and which ones create six new problems while solving one). I use AI in my writing because I’m interested in what it can make better without letting it flatten the parts that sound like me.

That, to me, is the most useful way to think about AI right now. You don’t have to overhaul the way you work or become the person posting breathlessly about 47 new AI tools every Tuesday. Pick something you actually need to do and see if AI can help you do it better.

Sometimes it will. Sometimes it absolutely will not. And sometimes you’ll spend an entire workday building a website and emerge weirdly delighted by the whole thing.✨

✏️Your 20-Minute Homework Challenge

Set a timer for 20 minutes and look through the things you’ve created or shared at work over the past month: decks, PDFs, briefing documents, spreadsheets, status reports, planning docs, whatever.

Pick one that might actually work better as an interactive site. Then give it to ChatGPT Sites with a simple challenge:

“Turn this into a site that makes this information easier to understand, use, and share. Don’t simply recreate the document, rethink how the information should work in this format.”

Give yourself 20 minutes to play with what it creates, then iterating/tweaking/getting it where you want it.

You Beta Believe It is a newsletter for communications professionals mastering AI and integrating new technologies. Enjoyed this issue or learned something new? Please pass it along!

I recently heard someone explain why the term GEO (Generative Engine Optimization) is a little misleading, and it got me thinking.

It’s tempting to view GEO as just a fresh spin on SEO (Search Engine Optimization)—another item to tack onto the marketing stack. But GEO doesn’t operate like that. SEO has a clear owner and a dedicated budget line, sitting safely within content and web teams focused on keywords, site performance, and responsiveness.

But GEO touches almost everything.

It’s your website, sure. But it’s also your executives’ LinkedIn posts, and the Reddit thread where someone asked for a recommendation in your category. It’s the trade publication featuring your CEO that once seemed too niche to matter, and the Amazon review from eighteen months ago. It’s the recent conference panel transcript, and the podcast interview nobody remembers, and the mention on Instagram, the complaint on X, and the product demonstration on YouTube. Every single bit of this is now source material for AI systems to digest, interpret, and describe your business to prospective customers.

And, with what we call “search” largely becoming AI-enable, (as in bot traffic has officially surpassed human traffic 😳), what an LLM says about you depends on the full depth and quality of your digital footprint, not just the tidy pages you control. This is why GEO can’t be tucked into one team. No single department owns 'everything published about us across the internet.'

I laid out the case for taking this seriously back in June—if you missed it, start there for the why. Today’s post is about the how and the what—as in, what these tools can actually do for you, and how to bridge the gap when they fall short.

SO. MUCH.CHANGE in just Two Months🤯

The speed at which this category is evolving is kind of wild.

Since I wrote about this topic on June 5th—just a little over two months ago—Profound, one of the AI visibility platforms I mentioned, has already added a 360-degree agentic execution layer with their new Aim and FactCheck products. So, it doesn't just flag when your brand is misrepresented anymore; they can actually act on it. It has officially morphed from a passive monitoring tool into a full-blown remediation platform.This is a meaningful shift. We’re now moving from “tell me how AI sees my brand” toward “tell me what’s wrong and help me do something about it.”

The catch, of course, is resources. 💵

Profound’s self-serve plans run from $99 to $399 a month (with custom enterprise pricing tiers) and for a tool that monitors and now actively cleans up AI messes so you don't have to, that pricing actually makes a ton of sense. It’s doing heavy lifting. But if you’re at a startup, a mid-sized brand where three departments are fighting for marketing dollars, or you are the entire comms department, spending enterprise cash on fancy GEO software probably isn't getting approved this quarter.

And, that’s a lot of us.

Which means the answer isn’t necessarily “buy the platform.” (Although, don’t we all love it when that happens 😏). It’s: build the discipline the platform would have given you, using what you’ve already got.

🛠️ Build your own AI Citation Dashboard (and guess?! I already did and you can copy it)

To make this work, you need a system to operationalize GEO—not another strategic document gathering digital dust. You need to track your top prompts, capture a real baseline, assign who owns what (even if “the team” is just you in a trench coat), and systematically target the specific sources shaping AI answers in your industry. Over and over again, as a routine.

So, I built a tool to help (using AI, of course)! It’s a Notion template called the AI Citation Dashboard, and you can duplicate it into your own workspace right now. 🎁

It does four things:

• Tracks your priority prompts: the handful of questions worth checking regularly, not just once.

• Captures your baseline: what the AI said, how accurate it was and what it cited.

• Assigns ownership: PR, Marketing, Content/SEO or cross-functional, so this doesn’t quietly become nobody’s job.

• Tracks your citation-source strategy: the specific listicle, Reddit community, trade publication or other source you’re targeting, along with where that effort actually stands.

And now, for the “Get-It-Done” Playbook, Function and Owner

This is where GEO transitions from a strategic concept into an everyday execution problem. I’ve taken the sharpest GEO tactics from my master list and sorted them by function. Why? Because "cross-functional responsibility" has a sneaky habit of turning into "zero accountability."

👉 What PR owns

Build a rapid-response bench. Have a short list of approved spokespeople, their subjects, and pre-cleared background information ready before news breaks. Lead with the insight, not a paragraph of executive biography. Reporters remember the person who was useful when they needed someone.

Protect the coverage you’ve already earned. When something is published, carefully review the company description, category labels, product names, titles, and statistics. Request corrections when facts are wrong and maintain one source-of-truth fact sheet so that everyone representing the brand is working from the same information.

Extend its life. Be sure to add earned coverage to your website newsroom. Link to it from relevant pages. Reference it in future pitches when appropriate. Connect that third-party validation to the rest of your digital footprint instead of letting it disappear after the initial hit.

Pursue awards that actually mean something. Prioritize awards with legitimate judging criteria and credibility in your industry. “Award-winning” isn’t especially useful if nobody respects the award.

👉 What Marketing owns

Turn what you already know into something citeable. Original research, customer surveys, benchmarks, pricing analyses … anything specific and defensible, backed by a clear methodology works. Give other people a reason to cite you.

Build campaigns around one genuinely interesting finding. Find the statistic that makes someone stop scrolling and build several smart angles around it rather than spreading ten mediocre findings across fifty low-value sites.

Build partnerships with substance. Joint research, co-authored reports, real collaborations (not just a logo swap).” Structure the work so every partner has a legitimate reason to talk about it, publish it, and link to it.

👉 What Content/SEO owns

Clean up your listings. Directories, review platforms, aggregators: claim your profiles, fix the categories, and keep the information consistent and current. It’s deeply unsexy work, but it’s also cheap to fix, and you’d be amazed how much outdated information is quietly sitting around the internet.

Measure quality, not just volume. Raw mention count doesn’t tell you enough. Ask better questions: Is this mention about a topic we actually want to own? Is the source credible with our audience? Track mentions, retrieval, citations and recommendations separately because they’re four different outcomes.

👉 What’s genuinely cross-functional

Participate in communities like an actual person. Reddit and similar communities increasingly matter because they’re places where people exchange detailed, candid opinions about products and categories, and that content can surface in AI-generated answers. But please don’t turn this into “our intern should go pretend to be a customer on Reddit.” Show up. Answer real questions. Be useful. Disclose who you are. Manufactured conversations and fake advocacy undermine the exact thing you’re trying to build: credibility.

The Bottom Line

GEO doesn’t fail because the tactics are hard. GEO fails because it has no owner, and work without an owner gets dropped the moment an inbox catches fire. So grab this free tool, set up your AI Citation Dashboard, and don’t leave this post with simply a better concept of GEO. Leave it with one Reddit thread claimed, one listing fixed, and one earned hit linked back to your newsroom. Small, assigned, and done beats smart, hypothetical, and someone else’s problem every time.

Subscribe now

✏️ Your 20-Minute Homework Challenge

Open and copy the dashboard I made(link here) and add one action for each function on your team: one PR action, one Marketing action, one Content/SEO action.

Don’t overthink it. Pick the tactic from each section that feels most doable this week, assign it to an actual human being, and give it a concrete next step.

The goal isn’t to leave this post understanding GEO better, it’s to do the actual work. You can make your brand’s AI footprint slightly better than it was just twenty minutes ago! 💪

Leave a comment

You Beta Believe It is a newsletter for communications professionals mastering AI and integrating new technologies. Enjoyed this issue or learned something new? Please pass it along!

Share

Hello again! 👋🏻

So, I did not intend to take the month of July off from writing this Substack, but somewhere between the lazy mornings, long walks with my kids, and celebrating a milestone birthday, I found myself craving moments without striving, hustling, optimizing, or learning the next thing. It was sooo nice.

This summer also shifted my work life in ways I didn’t expect, and I know I’m far from alone in adjusting to a workplace being reshaped by layoffs, downsizing, hiring freezes, and AI-driven change.

I’m sharing this to say that in our quest to learn new technologies and grow our careers, don’t forget to savor the real life moments along the way. They are the treasures that sustain us when it feels like the world is spinning just too fast (and it sure as heck is!)

Now let’s get into it … While I took July off, AI did not. 🤖

In the spirit of my occasional AI News That Matters installments, here are four developments I’m watching and what they mean for us in comms.

1. AI is now a geopolitical issue, whether we like it or not

I know. “Geopolitics” is such a cable-news chyron word, and it feels completely abstract to our daily work lives. But stick with me, because six months from now this is the thing your CEO asks you about five minutes before a board call.

The U.S. spent July drawing sharper lines around powerful AI: Chinese competitors, suspected copying, the future of open models, all of it. Axios covered the shift here.

👀 Why it Matters for Comms

Here’s why that matters to us: once a government starts treating AI models like strategic assets (something to protect, control, compete over, or keep out of certain hands), the tools your organization runs on stop being just software choices. They turn into questions about access, vendors, data, and trust.

What happens if the tool your team built into its workflow changes its rules overnight, gets pulled from a region, or gets swept up in a policy fight? Or your CEO spends six months publicly bragging about your company’s “AI advantage,” and a reporter asks which model powers it, what data it touches, and what guardrails are actually behind it?

👉🏻 What to consider

For people in our world, the practical takeaway is to know the answers to these questions:

• Which provider are we depending on?

• What happens if access or terms change?

• What can we accurately claim about how this tool works?

• Do our public promises match the actual guardrails behind the scenes?

  Thanks for reading You Beta Believe It ! Subscribe for free to receive new posts and support my work.

  ---

2. AI is becoming the front door to the internet

In June, I wrote about why PR is becoming more important in AI search. That core premise hasn’t changed, but the conversation around it has gotten a lot louder. During Google’s Q2 2026 earnings call, CEO Sundar Pichai reported that AI Mode had passed one billion monthly users and was being folded into AI Overviews for a more seamless search experience.

👀 Why it Matters for Comms

With bot traffic now exceeding human traffic, the question isn’t just “How do we show up in search?” anymore. It’s “How do we stay accurate, credible, and worth seeking out when an AI is assembling the first answer about us?” The new reality of brand representation is mastering how AI systems interpret your brand. Since these models pull straight from what’s already published about you online, brands need to get proactive: figure out which sources are currently shaping your category’s narrative, and get intentional about what you want AI systems to associate your brand with, and what you don’t.

👉🏻 What to consider

For communicators, that means the core work we’ve always done matters in a whole new way:

• Continue building credible third-party validation (earned media, for example)

• Maintain a crystal-clear corporate narrative

• Correct outdated online information

• Equip subject matter experts with genuinely useful insights

3. The AI coworker went mainstream while I was away

Before taking July off, I was deep in my Codex testing era, learning firsthand that it wasn’t just another chatbot. It was an actual coworker. I gave it a goal, context, and permission to handle multi-step work while I went and lived my life.

And while I was away, the rest of the industry caught up. Anthropic expanded Claude Cowork to web and mobile, and OpenAI answered with ChatGPT Work, powered by GPT-5.6. Both move beyond single prompts into executing full workflows: pulling context across connected tools to build documents, spreadsheets, and reports, all while keeping you in control.

We’re also watching this ecosystem consolidate in real time. The Codex desktop app and ChatGPT are beginning to converge, though the rollout is happening in stages. If you’ve accessed the newer Codex experience, you may already see Work inside that app, even if your standard ChatGPT interface hasn’t caught up yet. OpenAI is building them as two sides of the same desktop experience, but access is still trickling out account by account.

Ultimately, the important shift isn’t the branding. It’s the workflow. We’re officially moving from “ask AI a question and copy-paste the answer” to “give AI an outcome, source material, and boundaries, then review the first draft.”

👀 Why it Matters for Comms

For communications teams drowning in repetitive assembly work (coverage reports, media lists, briefing docs, reformatting one announcement into 17 different platform sizes), that’s a real gift back of your time.

👉🏻 What to consider

The essential skill for communicators is no longer prompt engineering. It’s delegation: knowing what to hand off, what context to provide, where sensitive data lives, and precisely where human judgment has to stay in control.

4. AI disclosure is becoming a brand practice

In July, Google rolled out a new “How this ad was made” panel across Search, YouTube, and Discover, letting people see when an ad was created or edited with AI. That might sound like a small platform update, but it’s actually a preview of a much bigger shift (IMO).

👀 Why it Matters for Comms

AI is moving from behind the scenes to right in front of the people we’re trying to reach. Customers will increasingly know (or reasonably assume) that AI helped make the image, write the copy, power the chatbot, recommend the product, or answer their question. Whether they care will depend less on the technology itself than on whether the use feels honest, useful, and aligned with what the brand claims to value. Which is exactly why disclosure is a communications question, not just a legal one.

👉🏻 What to consider

Not every AI-assisted sentence needs a giant warning label, but someone inside the organization needs to decide when AI involvement would materially change how a reasonable person understands what they’re seeing, hearing, or being asked to trust.

The Bottom Line

So there you have it, that’s the catch-up. The biggest shift I’m noticing is that none of this is really about the tools anymore , it’s who’s behind them, how people are finding information now, and whether we can explain any of our own choices when someone actually asks.

I’ve also come to the realization that staying on top of the bigger picture matters just as much as actually using the tools. It’s all too easy to focus on what’s in front of us — the tools we’ve been handed, the tasks we’re using them for — and forget to zoom out and see where the industry’s actually headed. But that’s kind of the whole job, isn’t it? Not just working the tools, but staying close enough to the bigger picture that we can help our companies stay ahead of it, and build the responses and policies before we need them instead of scrambling after.

✏️ Your 5-Minute Challenge

Want to sound like you actually know what’s happening in AI at your next team meeting? Use this roundup as your cheat sheet — steal a stat, drop a line about geopolitics, sound like you’ve been paying attention the whole time. Then go enjoy the last few days of summer. I highly recommend it! ☀️

Leave a comment

You Beta Believe It is a newsletter for communications professionals mastering AI and integrating new technologies. Enjoyed this issue or learned something new? Please pass it along!

Oh holy Codex … where do I start.

First, let me say this plainly: I am at the beginning of learning this tool. This is not a “here is my flawless expert system for running my entire life through AI” post. It is much more of a “I opened the thing, it broke my brain a little, but I kept tinkering and made it do some genuinely useful work” kind of post.

I continuously refer to AI as a magic act, because so often, it is (not to negate the very real and catastrophic environmental costs of it) To have technology that can connect to your email, your files, your phone, and then just whip up a website, a dashboard, a summary, a presentation — based on a few simple sentences you type, in minutes — is genuinely staggering. 🤯

But before I get into what I actually did with it, let’s back up. Because the most important thing to understand about Codex is that it is not a chatbot.

So, What Actually Is Codex?

The plain-English version: Codex is OpenAI's desktop AI agent. It started as a developer tool, but OpenAI has been expanding it for the non-engineers, the knowledge workers, the people who just need things handled. While it looks a lot like ChatGPT, and can do many of the same tasks, it’s also quite different because it can work across your computer, use plugins, connect with apps, generate images, read files, build things, and run automations. It is not just a chatbot sitting in a browser waiting for you to say something clever, think of it more like a coworker, or a very capable buddy, or, depending on how many permissions you hand over, an overbearing mother who looks at your mess and says, "Fine. I'll handle it."

Now, that does not mean it should act without supervision. Dear God, no. But it does mean the mental model is different. With ChatGPT, I usually ask for help with a single piece of work. With Codex, I can ask it to produce the thing, organize the folder, draft the email, build the dashboard, make the spreadsheet, create the visual, summarize the morning, or prepare the briefing. And, while you watch it happen inside the “terminal” (more on that below), it’s mind-boggling.

A few important things to know upfront:

• It’s free with a paid ChatGPT account. If you have ChatGPT Plus, Team, or Enterprise, Codex is already included in your subscription.

• It’s a desktop app, not a browser tab. You use your ChatGPT login to sign in and then you download it. It lives on your computer, which is what allows it to interact with your local files in ways browser-based tools can’t.

• It’s powered by GPT-5.5. OpenAI’s latest model is widely regarded as the strongest all-around AI model available as of this writing. It is optimized specifically for agentic work like this.

• It’s genuinely different from a chatbot. Sorry to be repetitive here, but truly understanding this distinction is important.

🫣 The AI Block is Real

Here is where I think a lot of us get stuck.

We hear the AI crowd waving their hands enthusiastically saying, “Codex can do everything! It can run your workflows! It can use your tools! It can build whatever you imagine!”

And then we sit down at our desks, open the app, stare at the blinking prompt, and think:

Okay … but what is my grand idea?

What exactly do I need it to do?

This has been one of the biggest surprises of my AI education. The barrier is not always the tool. Sometimes the barrier is imagination under pressure, or as I call AI Block. It’s the strange paralysis that happens when a tool can do so much that you suddenly cannot think of one useful thing to ask it. It’s like being handed the keys to a commercial kitchen when all you wanted was lunch.

For me, the best way around that block has been watching other people use the tool—as in pulling a few You Tube videos and watching. Even if the use case has nothing to do with my life or my work, seeing someone else’s prompt opens a little door in my brain. Oh, wait. I do have something like that. Oh, wait. I do waste time on that. Oh, wait, that thing I keep manually rebuilding every week could probably become a repeatable workflow.

And this is where Codex gets really interesting for communications professionals.

🧐 Why Communicators Should Pay Attention

Communications work is full of judgment, nuance, relationship management, institutional memory, audience sensitivity, and context.

It is also full of repetitive assembly work. Pulling coverage, building briefing docs, scanning headlines, drafting first-pass responses, creating content calendars, organizing files, repurposing event materials into five different formats because apparently every channel needs its own precise little rectangle of content, preparing executives before interviews, reading long threads and figuring out what actually matters, following up on the things everyone agreed to but nobody wrote down, and on and on. And, this is where Codex starts to look less like a shiny AI toy and more like a serious operating layer for knowledge work.

I recently attended a private Codex demonstration specifically for communications professionals (many thanks to OpenAI’s Lindsay McCallum Rémy for making that happen) and the use cases were legitimately impressive.

A few that stuck with me:

• “Chief of Staff” morning briefing: Set it up once and it checks your Slack, Google Calendar, and Gmail every morning, drafts a priority summary, and delivers it to your preferred Slack thread by 8am. No “@” command needed — just tell it in plain language what you want checked, and if the plugin is enabled, it knows where to look.

• Email triage and response: Flag the most urgent emails requiring your attention, draft replies in your voice, and either send them or surface them for your review first.

• Executive interview prep: Build a Reporter Briefing document using the “Sites” function before a media interview, pulling from files you already have on hand.

  • Sites (worth knowing about, even if I can’t fully test it yet): One feature that got my attention during the demo is called Sites — a capability inside Codex (Business and Enterprise accounts only for now) that lets you build a living, shareable microsite that your team can access and update dynamically.

• Event logistics: Create a seating chart visual for a media dinner that you can drag and drop guests at the table as RSVPs roll in.

• Content calendar: Reference your existing brand files and build out a content calendar directly into a Google Spreadsheet.

• Promotional materials: Generate visual and copy assets for an event, sized for multiple placements — television display, social posts, flyers — using a preferred style guide.

• Daily news monitoring: Scan relevant headlines on a schedule and deliver a briefing to you by 9am.

  **Dear Reader, you may remember that I spent a non-trivial amount of time trying to build my own AI news briefing system with a custom GPT and Zapier. Seeing Codex do a version of that made my earlier workflow feel a little like showing up to a Tesla dealership on a unicycle. 😑

  ---

  🛠️ What I Actually Did With It

  So, I didn’t hand Codex my entire professional life and say, “Good luck, Bestie.” Rather, I started smaller, because that is usually the only way I learn without completely spiraling.

  I had it do a few things:

  1. (Work) Built a PR coverage dashboard. I fed it data from a recent company announcement and asked it to build a dashboard to visualize the coverage.

  2. (Work) Organized a chaotic work folder. I had 100+ disorganized folders and files on my desktop that had been driving me crazy for months. I asked Codex to sort through one of the particularly bothersome folders and organize all of it for me into subfolders based on content.

  3. (Personal) Pulled family-friendly dinner recipes from a website I love. I asked it to visit a favorite cooking website (shout out to and Jenn Eats Good) and pull only the dinner recipes, filtering for things that would survive contact with two extremely picky kids.

     Within minutes, it sorted through the 240 dinner recipes on the site to find 62 saved as “family fit” options. It then reviewed, grouped, and rated each of these 62 recipes according to “Best Bets,” “Solid Weeknight Options,” and “Maybe with Modifcations” categories. From there, it indexed all of them into one document and created a spreadsheet of the recipes in both a sortable fashion and a complete list of all 62 dinner recipes. And, as if that wasn’t enough, as a cherry on top, per my instructions, it then create an organized Google drive folder of all of it for easy access. It is a thing of beauty. 🙌🏻

  4. (Personal) Sent in a prescription refill request on my behalf. I’d been putting off emailing my son’s doctor about a prescription refill, and in between errands on a busy day, I used the mobile function of Codex to draft and send an email requesting a refill request to his doctor, who is accessible through my contacts. The refill was called in all while I was out and about, easy breezy!

The dashboard example was the one that made my comms brain light up first because coverage reports are one of those things that can be either deeply useful or spiritually punishing depending on the format, deadline, and number of screenshots involved. Codex was able to take source material, organize it, and turn it into something more useful and visually-appealing than a pile of links.

🤓 The Vocabulary Breakdown (Because it can be Confusing at First)

When you first open Codex, you’ll hit some terms that aren’t immediately obvious. Here’s the plain-English translation:

• Skills: Reusable workflows or capabilities. If you repeatedly do the same multi-step thing, you can build a Skill so you don’t have to re-explain it every time. There are also some pre-installed Skills available out of the box.

• Plugins: The connections to your external tools — Gmail, Slack, HubSpot, etc. This is where you give Codex actual access to work inside your other apps.

• Automations: Instructions for “when” to do something such as “Every weekday at 8am, check my email and send me a briefing.”

• Projects: A way to organize work around a specific goal, using your actual computer files and folders as the source material.

• Terminal: The text-based control panel for more advanced tasks.

• Harness: The defined sandbox space where Codex is allowed to operate. (More on this in the safety section below).

The important thing is that these pieces work together. Skills teach it how. Plugins tell it where it can work. Automations tell it when. Projects give it context. Permissions tell it how much freedom it has.

⭐ Bonus Tips:

• If you are inside Codex and want to see extra options, type `/` in the prompt bar. A menu appears with all kinds of skills, tools, and option.It is basically an instant grocery list of capabilities, and I wish I had known it existed earlier.

• Also, the plugins/skills area confused me at first. Plugins show up in the sidebar, and from there you can toggle into skills. Once I found it, it made more sense.

• The `/pet` command is also delightful. If Codex is working on something that may take a while, it can create a little floating status companion in the lower right corner of your screen so you can keep working elsewhere while it updates you or asks for input.

The Snags (Because There Were Several)

You know I’m going to tell you when things didn’t work.

• The mobile setup was genuinely a whole thing. One of the cooler Codex capabilities is that you can connect to it from your phone. So, you can be away from your desk and still ask it to do work based on what you’ve connected. In order to enable this, you have to set up multi-factor authentication first and that process sent me down a rabbit hole that I was not prepared for on a Tuesday afternoon. I ended up using the Authy app to sort it out, which worked, but the whole thing was not fast or intuitive (for me, at least).

• Mobile requires your computer to be awake. I learned this while driving to North Carolina to pick up my daughter from camp, when I tried to use Codex from my phone and nothing happened. My laptop was idle at home. It seems obvious in retrospect, but happy to share my foibles in case it helps you.

• The plugin/skill navigation is not intuitive at first. I noted this above but reiterating here. Plugins appear in the sidebar and skills are found under Plugins. Don’t go on a egg hunt looking for them as I did.

  ---

  🦺 A Word on Safety (Please Read this Part)

  Which brings me to the part where we need to talk about safety. Giving an AI agent access to your email, your contacts, your files, and your calendar is not a casual decision.

  Should we worry about sharing so much information with an agentic tool? Yes, absolutely. Any technology that can access this level of your information deserves a healthy amount of suspicion. Not panic or avoidance, but serious caution.

  This is especially true for communications professionals because we often sit near sensitive information: executive plans, employee issues, legal reviews, confidential announcements, customer stories, media strategy, crisis scenarios, board materials, and all the other things that should not be casually pasted into a tool.

A few practical guardrails I’m using and recommend:

• In Codex Settings > General Permissions: Do not enable “full access.” Limit what it can see.

• In Codex Settings > Configuration > Approval Policy: Enable “untrusted.” This means Codex will always ask your approval before taking action. Adds a step, and it’s worth it.

• In ChatGPT Settings > Apps > App Preferences: Find “ask permission” and set it to always ask.

• Create a Sandbox folder: Make a new, isolated folder on your desktop — name it something like “Codex Testing” — and use that as the designated space for files you want Codex to access. This keeps it from scanning everything on your computer.

• Treat every output as a draft until a human has reviewed it.

That last one really matters.

Codex can do a lot, but it cannot take accountability

If it sends the wrong email, moves the wrong file, misreads the tone of a reporter note, or produces a briefing with a missing fact, that is still on you.

The Bottom Line

Codex is not a shinier chatbot. It’s a different category of tool entirely, and once you understand that distinction, things start to click.

The tasks that used to eat a significant chunk of your day are now things you can describe in a sentence and walk away from. And, the real skill is not prompting anymore. Just a little while ago, knowing how to write a decent prompt felt like a competitive advantage. Now, that is table stakes.

The next skill is learning how to delegate to AI systems, which sounds simple, but it’s not. Delegating to a person requires clarity, context, expectations, feedback, and trust. Delegating to AI requires all of that, plus guardrails, permissions, review paths, and a very clear understanding of what good looks like.

• You have to know the work well enough to break it apart.

• You have to decide what can be handed off and what cannot.

• You have to define what “done” means.

• You have to carefully inspect the result.

• You have to build enough confidence to try again and again until you get it right.

This is why Codex feels so powerful and so intimidating. It asks us to stop thinking of AI as a place we go for answers and start thinking of it as something we manage, which is a big shift.

Codex isn’t perfect, but I do think it’s one of the clearest signals yet for where work is going. And, as intimidating as Codex (or any agentic work can seem,) I would rather be the person learning it now, than the person pretending the shift isn’t happening.

✏️Your 20-Minute Homework

Pick one thing on your desk right now that you’ve been avoiding. Maybe it’s the folder of files that needs organizing, or the email you’ve been drafting in your head for three days, or the briefing doc you keep meaning to put together.

Download the Codex desktop app, connect one plugin you actually use (Gmail is the easiest starting point), and ask it to handle that one thing.

Don’t try to build a daily morning briefing automation on day one. Don’t connect six plugins. Just one task, one plugin, twenty minutes.

Then come back and tell me how it went in the comments.

Leave a comment

You Beta Believe It is a newsletter for communications professionals mastering AI and integrating new technologies. Enjoyed this issue or learned something new? Please pass it along!

Share

Back in April 2025, Emily Ley (yes, the same one who writes the #1 parenting substack, The Landing, who is also the Founder of Simplified® and author of multiple bestselling books) and I had just landed in Orlando for her Mom 2.0 keynote when both of our phones exploded simultaneously.

Her lawsuit had been officially filed: Emily Ley Paper, Inc. v. President Donald J. Trump—marking the first civil lawsuit challenging the constitutionality of the Trump administration’s worldwide tariffs. Overnight, Emily went from lifestyle guru to international news icon, and we were immediately thrust onto a wild roller coaster together. Her name was everywhere. CNN wanted her. MSNBC. BBC. ABC. CBS News. CBS Mornings. The Washington Post. Fast Company. Reuters. Bloomberg. The requests just kept coming.

There we were, holed up in a hotel room. She had forgotten her laptop, and mine didn’t have a camera that met broadcast quality standards. So, naturally, we DoorDashed one. 🤡

And, in a beautifully chaotic way, this story illustrates a career in communications: you plan, you prepare, you pitch, and then something completely unexpected happens and you figure it out with whatever you have … even if that means a DoorDashed webcam.

But, let me back up.

At the start of 2025, Emily—whom I’d worked with ten years prior and remained great friends with—reached out with a simple, semi-desperate ask: Did I still do book publicity? Technically, not recently. My first job out of college was working as a publicist for a small book publisher in Atlanta. From there, I moved to New York, spending many years working in publicity at Doubleday and then Hyperion Books. And, while my career has taken plenty of turns since, I’ve never really left publishing behind. It will always have my heart.

So when Emily called, I said yes. She needed immediate PR support for The Simplified Cookbook. We had very little runway, so I quickly went to work—pitching lifestyle and food media, lining up her book tour, and generally sprinting to catch up. Then, Emily shared a candid, factual social media post about the devastating financial impact tariffs were having on her small business, and a Washington, D.C.-based civil rights organization saw it and came calling asking her to take a very public, legal stand against it all. The rest was, well … history.

I held a front-row seat to something genuinely historic. And watching it unfold—watching Emily absorb the scrutiny, stand in her conviction, do interviews from a hotel room with a camera that arrived in a little orange bag—it was deeply meaningful.

This week, the experience came full-circle (more on that below), and it’s gotten me thinking about everything I know about this work … the things I’ve learned slowly, sometimes the hard way, and across two-plus decades.

So here’s what I know for sure:

You can’t sell a person or product that isn’t good. Sure, you can put a lot of resources behind something mediocre and get some initial noise. But the public is smart and word-of-mouth is undefeated. Eventually the campaign runs out of steam and the thing just sits there.

The “big” placement isn’t always the right one. Everyone wants The Today Show, The New York Times, or The Wall Street Journal. And yes, those matter. But more often than not, a niche podcast interview or a targeted industry speaking engagement can move the needle for a client far more than a legacy media holy grail (not to mention what it can do for AI visibility).

Relatability is the whole game. You can have the most exciting media opportunity in the world and completely waste it if you can’t succinctly convey your point of view and make the audience feel an immediate connection. Trust me, this is much harder than it sounds. Authentic presence can’t be faked—you can practice and you can train, but true on-camera authenticity is a gift.

Glamour is a cover story. When I started working at OWN:The Oprah Winfrey Network as a member of the launch team, we worked out of a dilapidated office building allllllll the way out near Bob Hope Airport in the heart of Burbank (IYKYK). Think greige walls, makeshift cubicles, and one shared printer. We were building, from scratch, one of the most anticipated cable television launches in years, and the physical reality was ... not Oprah glam. Every “exciting” industry involves hard work, long hours, and, oftentimes, very unglamorous settings.

Your clients are trusting you with their baby. Whatever someone is putting out into the world — a book, a lawsuit, a brand, an opinion — it’s theirs. They are (carefully and vulnerably) handing it to you. Treat it accordingly. Today’s media environment is loud, fast, and not forgiving. They are counting on you to shepherd them through it with care, not just competence.

Your media relationships are a two-way street. The journalists, producers, and hosts you work with are counting on you, too. They need you to prepare your client, deliver a clean experience, and not waste their time. Relationships and basic human decency go a long way in this industry.

Campaigns go sideways. That’s the job. Even the most carefully crafted plan can fall apart, take a hard left turn, or completely explode. Sometimes it’s your fault. Sometimes it’s just the news cycle. The job is to figure it out — to pivot, to fan the flames, to take accountability. Emily didn’t plan to become an international news story in the middle of a cookbook tour. But, that’s what happened, and we figured it out.

You have to keep evolving, or you become a cautionary tale. A long time ago, a (very short-sighted) boss told a room full of publicists not to worry about bloggers, because “that trend won’t last.” 😳 Now with AI disrupting every part of our industry, the people who thrive are the ones who stay curious and keep learning — which is, not coincidentally, the entire reason I write this newsletter.

Yesterday, Emily invited me to Washington, D.C., where she accepted an award at the Library of Congress for the courage it took to stand up and speak out publicly on behalf of small business owners when it would have been much easier to stay quiet.

As I stood there in Madison Hall—surrounded by the poignant inscriptions of James Madison’s speeches and a striking marble statue of our fourth U.S. President—I felt something I never take for granted: a deep gratitude for the small part I played. That the camera we DoorDashed in haste to a hotel room in Orlando had turned into something that truly mattered. After 25 years, there is still no better feeling.

⭐ My final thoughts: If you’re putting yourself out there — with your book, your company, your opinion, your newsletter — don’t ever let fear or vanity get in the way of your mission.

What’s a moment in your career where something went sideways and you had to figure it out on the fly?

Leave a comment

You Beta Believe It is a newsletter for communications professionals mastering AI and integrating new technologies. Enjoyed this issue or learned something new? Please pass it along!

You know this moment.

You land a great story in a respected trade publication. The audience is exactly right. The publication has credibility. The readers are buyers, decision-makers, and industry insiders.

You proudly send the coverage to the executive team. The CEO glances at it and says, “Nobody reads this. Can we get something in The Wall Street Journal?”

Cue the internal screaming.

Communications professionals have been living this conversation for years.

The podcast you pitched? Too niche.

The Substack feature? Not enough subscribers.

The industry newsletter? Too small.

The YouTube channel? Never heard of it.

For decades, the value of these placements was difficult to quantify in a way that resonated with the C-suite. We knew they mattered, that they reached the right people, that credibility isn’t always measured by audience size, but what we couldn’t fully explain was that these seemingly “small” hits were building something much bigger.

Now we can. It’s called Generative Engine Optimization, or GEO, and it may be the best thing that’s happened to public relations in decades.

As AI increasingly becomes the way people discover companies, products, experts, and ideas, many of those niche publications, podcasts, newsletters, bylines, and trade media placements are becoming part of the information layer that trains and informs generative search engines.

In other words, that trade publication your CEO dismissed may have more influence on whether ChatGPT recommends your company than a one-off national media hit.

In fact, here is what’s actually happening while your leadership team is protecting their calendars waiting for that major national feature that may never come:

• YouTube is now cited in 16% of LLM (Large Language Model) answers — officially beating out Reddit as a primary source of truth (Bluefish, Jan 2026, reported by Adweek).

• LinkedIn is the #2 most cited domain across ChatGPT, Google, and Perplexity (Semrush, 325,000 prompts analyzed, Feb 2026).

• Substack is being crawled continuously by AI systems as a high-trust, high-engagement editorial source (PR Daily).

• Freshness matters: Content updated within 30 days gets 3.2x more AI citations than older content.

The podcast with “only” 2,000 listeners? Its transcript is indexed and its audio is parsed. AI systems love it. And, the vertical newsletter mentions your client waved off? They are one of the strongest authority signals AI tools use to decide who gets recommended in your category.

GEO vs. SEO: Same Alphabet, Very Different Game

Traditional SEO is about getting your website to rank high in Google search results so people click through to your site. The whole game is about driving traffic.

GEO is something entirely different. It’s about making sure that when ChatGPT, Claude, Perplexity, or Google Gemini answers a question in your category, your brand gets cited as part of that answer. SEO gets you clicks. GEO gets you citations. The user might never see your website. The AI just recommends you based on the data it trusts.

Here’s why that distinction matters right now: experts predict traditional search will lose 50% of its market share by 2028 as users migrate to conversational AI. Furthermore, data shows that buyers who do discover a brand through AI search convert at 4.4 times the rate (!) of traditional organic visitors. So, while the volume is shifting, the quality of the audience is skyrocketing.

The catch? You can’t buy your way into ChatGPT. There is no pay-per-click model for an LLM answer. Up to 90% of citations that drive brand visibility in AI responses come from earned media — third-party editorial sources, journalist-reviewed articles, and credible industry mentions.

Every media placement you’ve ever fought for, every reporter relationship you’ve cultivated, every byline you’ve placed in a trade pub — all of it is now the essential training data your prospects use to make buying decisions. AI engines don’t rank publications by subscription counts; they rank them by category authority. They are built to mirror how humans assess credibility: if a trusted independent party says it, it’s probably true.

Beyond just securing media placements, there are active, immediate steps communications leaders can take right now to deliberately optimize a brand’s AI visibility. It starts with knowing exactly where you stand.

1️⃣ Step One: Audit Your AI Presence

Before you can fix your positioning, you need to know what the AI is actually saying about you. Your brand might be placed in the wrong category, or your flagship product might be described as a legacy offering you retired two years ago.

💪🏻 Here’s a scrappy, basic way to run an AI brand audit:

1. Test the Big Three: Run your brand name, your product category, and your competitive differentiators through ChatGPT, Claude, and Perplexity. They pull from different source sets and will return completely different results.

2. Look for Erasure and Lag: Watch for brand erasure (not appearing in your own category- not good) and positioning lag (outdated corporate narratives- also not good).

3. Find What’s Being Cited: Dig into the footnotes. Know what channels, publications, and credibility sources are influencing LLMs in your specific market.

4. Document the Receipts: Screenshot any inaccuracies. This becomes your baseline data when you need to make the case internally for a dedicated GEO budget. Treat inaccuracies as a business risk, not a communications annoyance.

2️⃣ Step Two: Follow this basic GEO Playbook

Once you know where you stand, you don’t need to become a technical coding expert to fix it, you just need to evolve your existing PR practice to account for how machines evaluate credibility.

Here are your strategic priorities:

• Define Your Buyer-Centric Prompts: Identify the exact conversational prompts your target audience is typing into AI tools and build your media strategy around those queries.

  💡 Here’s one quick and easy way to do this: Ask your LLM with this prompt–

  "Act as a market researcher analyzing user behavior in AI search engines. My target audience is [Define Audience]. Provide a list of 15 highly specific, multi-turn prompts this persona would type into ChatGPT/Claude/Perplexity/Gemini when trying to solve [Specific Problem]. Include the context, constraints, and specific output formats they would ask for.

• Map the LLM-Trusted Landscape: Which outlets are AI systems in your specific vertical actually pulling from? Spend an hour testing prompts, studying what gets footnoted, and building an intentional target media list based on machine trust, not just human eyeballs.

• Elevate Your Executive Spokespeople: Train executives to lose the vague marketing fluff kind of speak—think, "We are leveraging next-generation, cutting-edge paradigms to synergize our client offerings." The AI’s algorithm flags this as high-noise, low-value information and skips right over it. Have them adopt the Context (the situation)-Insight (the expert opinion) -Implication (the real-world consquence) structure instead, as it forces the speaker to package their thoughts into a perfect, bite-sized data packet that AI layout engines love.

• Fix Positioning Lag Proactively: Don’t wait for a quarterly audit to catch outdated narratives. Coordinate your earned media push and structured data ahead of every product launch or corporate strategy shift to feed the models fresh data inputs immediately.

• Build Cross-Functional Accountability: PR, content marketing, and SEO can no longer operate in separate corporate silos. GEO requires a unified signal network, which means shared strategy, standardized terminology, and shared measurement frameworks.

  Thanks for reading You Beta Believe It! Subscribe for free to receive new posts and support my work.

3️⃣ Step Three: Update the Format of your Press Releases

The press release was already staging a comeback, but GEO makes it genuinely strategic again, provided you write it for machines as well as humans. AI systems place massive weight on structured, unambiguous third-party validation.

AI-optimized press releases pack everything important in the first 75-100 words, avoid vague terms (“platform” or “the company”), provide clear attribution (not “company spokesperson”), and follow a structured formatting. Additionally, the optimized headline formula is: [Organization] + [Action] + [Product/Service] + [Quantifiable Benefit].

4️⃣ Step Four: How to Measure AI Visibility

If you’ve been measuring PR success through vague impressions and clip counts, you already know those metrics carry less and less weight with leadership. GEO gives you a reason to upgrade the conversation to metrics the C-suite actually respects:

• AI Visibility Score: How prominently your brand appears in AI-generated answers for the specific prompts your target buyers are using. This is the new “page one ranking.”

• Answer Share of Voice (SOV): How often your brand appears in AI answers in your category compared directly to your competitors.

• Pipeline Influence: Connecting AI discovery directly to your CRM. With 59% of marketers already reporting measurable inbound growth from AI visibility, this is how you prove ROI.

If you want to move beyond manual audits, several specialized (and pricey!) platforms have emerged to track the path from prompt to recommendation. Generally-speaking, these tools provide comprehensive research (specific prompts/topics that buyers use in your AI search), daily tracking across AI platforms, audits and detailed recommendations for improvements (think tailored content pieces, Reddit threads to get included in, etc.), and more.

Here are a few of note (in random order):

• Semrush

• Profound

• Scrunch

• Meltwater

• Otterly

• AthenaHQ

• Limy

• Peec AI

  ---

The Bottom Line

Brands that treat PR as a nice-to-have function in the AI era are missing what may be one of the biggest shifts in discovery since the birth of Google.

As buyers increasingly turn to ChatGPT, Gemini, Claude, Perplexity, and AI Overviews for answers, the brands that get recommended won’t necessarily be the ones spending the most on advertising. They’ll be the brands that have built the strongest network of trust, authority, and credibility across the web.

Conveniently, that’s exactly what great PR has always done.😉

For years, communications teams have had to defend the value of earned media. In the GEO era, we may finally have the proof we’ve been looking for.

✏️ Your 20-Minute Homework

Open ChatGPT, Claude, and Perplexity and ask each of them these five questions:

1. What is [Your Company]?

2. Who are the leading companies in [your category]?

3. What makes [Your Company] different from competitors?

4. Who are the most influential executives in [your industry]?

5. What companies would you recommend for [problem your customers solve]?

Document the answers.

Then ask yourself:

• Did your company appear?

• Did your CEO appear?

• Was the description accurate?

• Were competitors mentioned more often than you?

• What sources were cited?

If the answers surprise you, you’ve just identified your first GEO opportunity.

Did your company show up? Did your competitors? And were you surprised by the answer? Let’s talk about it!

Leave a comment

You Beta Believe It is a newsletter for communications professionals about learning AI in public, in real time. If this one made you think or you enjoyed it, please share it!

And, watching him feels strangely emotional right now.

Maybe it’s because so much of the AI conversation this week has centered on backlash, anxiety, and what we may be losing as artificial intelligence removes more friction from how we work and create. Maybe it’s because even Pope Leo XIV is now warning us all about the slow erosion of the kind of human formation that happens when people struggle through problems, develop judgment, and learn by doing.

But, before I go full encyclical on you, let’s start with the fun stuff.

Disclaimer: This week’s post is a bit of a mishmash of experiments and news.

🧪 Beta Lab: I Vibe Coded an App, so yes, now I am a Developer.

I want to start with the thing I’m most genuinely excited about, which is also the thing that sounds the most like tech-bro nonsense: vibe coding.

Vibe coding, in the simplest possible terms, is building software using conversational terms instead of actual java/html/python computer code. You have a conversation with an AI tool about what you want using plain language, you then iterate through feedback, and end up with a working app.

I’ve never written a single line of code in my life, so I can’t tell you what this feels like compared to traditional development. But I can tell you that as a non-developer doing it, it’s kind of insane (like magic🪄)

Goal: Inspired by this “Mostly Human” podcast interview with tech investor Andrew Yeung where he coached host Laurie Segall through her first vibe coding session, I wanted to vibe code an app that will help with a particularly chaotic area of my life—my email inbox.

Tools: Lovable, Replit (for comparison)

The Experiment: Here was my original prompt:

“I am completely overwhelmed by my email inbox much of the time. I want something that can help me sort through what matters and also help me stay focused and on track to do the things I need to do each day. A schedule and routine that doesn’t get hijacked by my inbox.”

The Results: After several rounds of back-and-forth and testing, I created (or “shipped,” as they say) an online dashboard divided into four main sections:

• Today- provides a an overview of the day’s schedule, priorities, and emails that actually matter

• Triage- tackles my inbox and filters per level of importance

• Tasks- pulls important actions that are needed based on emails and those that I manually enter

• Tracker- creates a checklist of daily/monthly/quarterly habits tied to my goals for the year

I then connected it to my Google calendar and Gmail so it pulls in my actual appointments and data.

Lovable offers various color palettes to choose from for your work, and I landed on a bright multi-color scheme: green for “breaks,” blue for “focus,” purple for “inbox triage”), as well as a way to design and color-code your favicon (a yellow bullseye).

You’re able to then use the name they offer (Lovable suggested “Inbox Sanctuary”), but that felt too zen for me, so I went with “Focus Four.”

And, voilà, here it is: focusfour.lovable.app (which is public, hosted on Lovable’s site, and available to connect/tailor to your own needs), a hub for my everyday, in order to feel a little more focused and a little less stuck-in-my-overflowing-inbox-and-frozen in motion.

The whole build (meaning, all the iterations where I wanted this button added or that capability included or that thing isn’t working type of work) took maybe 3 hours spread across a few days (not consecutive).

Lovable vs. Replit: The Quick Verdict

I also tried Replit. Oddly, my favorite thing about working with Replit was silly but true. It makes these very satisfying whoosh/jingle sound after each iteration 🎶. But, beyond the audio experience, Replit was harder to talk to, harder to interpret, and just harder overall. So, I asked Gemini to summarize the difference and the description is accurate to my personal experience:

• Lovable: Best for SaaS prototypes, client dashboards, directory sites, or consumer apps where you want a polished, premium look with minimal design tweaking.

• Replit: Best when the backend logic is the whole point — heavy Python automation, custom server-side processing, or if you want to actually get your hands in the source code.

For me (and those like me), I’d say Lovable is your best starting point.

Pricing, for the record

Lovable offers free and paid plans across individual, business, and enterprise tiers. I’m on the Pro 1 monthly at $25/month, which gives you 150 credits and a private workspace. Credits fund your builds, in other words, the more complex your build, the more credits you use. The free tier is a solid way to test before committing.

📰 Switching Gears to The AI News That Matters

Similar to this post, I’ll frequently share a handful of the top AI-related news stories that are dominating the chatter and what they mean for comms teams.

1. Google Zero: The SEO Reckoning Has a Name Now

You may have already noticed the “AI Overview” that surfaces at the top of your Google search results. Ask it how to paint a wall and it will surface a step by step process with a few embedded links that you don’t need to click. This is not a coincidence, it’s a strategy.

At Google I/O last week, Google officially announced it’s transforming Search from a directory of links into an AI assistant, and publishers are terrified. The term circulating right now is “Google Zero” or the “looming death of search traffic.” It is a future where Google sends them zero traffic because users get the answer without ever clicking through.

Honestly, it feels unfair. Publishers create the content. AI learns from it. Users consume the summary. The website gets nothing for it.💸

From a communications standpoint, this is genuinely worth a hard look because:

• What does SEO even mean when the goal is to be quoted by AI rather than clicked on by humans (and yes, SEO vs. GEO is a future topic I plan to cover)

• If your client’s website is built for link clicks, is it still built for the right thing (the answer is no, and thus begins the work of rethinking everything 🫠)?

• Are the first two sentences of your “About” page doing enough work to stand alone as an answer (likely not and certainly not for the audible AI glasses, more on that👇🏻)

For more information on this, I recommend listening to The New York Times “Hard Fork” podcast interview with Google CEO Sundar Pichai. It’s an thoughtful (albeit biased) interview on this topic and on the broader AI disruption discussion.

2. Pope Leo XIV Would Like a Word

I did not expect to be writing about a papal encyclical this week. And yet …

Pope Leo XIV just released his first encyclical, “Magnifica humanitas: On Safeguarding the Human Person in the Time of Artificial Intelligence,” marking the 135th anniversary of Rerum novarum, the Church’s foundational document on labor and social justice.

A few things he’s worried about that hit close to home:

• Automation is compressing not just jobs, but the human formation process that creates expertise in the first place. When the foundational work disappears before younger generations can learn through doing it, what exactly are we passing on? (Yep, I feel you … I wrote about this recently.)

• Work is not only economic, it’s tied to identity, meaning, and contribution. AI systems that maximize productivity while hollowing out human participation create new forms of alienation.

• AI is shaping public opinion invisibly through feeds and recommendation systems, and this is worrisome, communications professionals know all too well.

He continues that we’re beginning to confuse:

• intelligence with wisdom

• information with truth

• automation with progress

• convenience with flourishing

His central question isn’t “Can we build this?”, it’s “What kind of people are we becoming while we build it?”

Whether you’re religious or not, it’s a serious and culturally relevant conversation to consider and one that has largely been dominated by investors and developers thus far. Which is also a problem 🤔.

3. Graduation Season & AI Backlash

At multiple commencement ceremonies this spring — Middle Tennessee State, University of Central Florida, University of Arizona — students have booed speakers who compared AI to the Industrial Revolution or the advent of the laptop. We’re talking audible, sustained booing.

One student described a speaker’s remarks as “a knife to the chest.”

A Cornell professor who studies technology adoption put it plainly: “These tech executives are not reading the room. These kids have spent hundreds of thousands of dollars on a degree that they don’t know will serve them well.”

The context that makes this sharper is that in 2025, the share of unemployed Americans who are new workforce entrants hit a 37-year high, peaking at 13.3% in July. That’s higher than at any point during the Great Recession. These aren’t just abstract fears.

What’s especially frustrating for these students is that most universities still have strict no-AI policies. So, they’ve spent four years being trained to not use the tools employers are now actively looking and expecting them to know how to adeptly use.

From a communications standpoint, if you’re advising a company or executive on AI messaging and you’re not pressure-testing it against the intended audience, you’re missing something real.

4. Google's AI Glasses Are Coming

Google just announced intelligent eyewear at I/O, shipping this fall. These are Gemini-powered eye glasses that speak one answer aloud when you ask a question. Not a list of links, no visuals, no bullet points, no sidebar, and no second screen. Just the AI reading one line out loud.

Let’s think about what this means for brand communications. The cues that corporate copy usually depends on — hierarchy, formatting, visual emphasis — won’t survive voice. If your client’s positioning statement is full of jargon, hedged language, and abstraction, it’s going to fail in the exact moment an AI chooses it as the answer to someone’s question.

What will perform in this environment is plain English, specificity, and a first sentence that can be quoted aloud and still make sense without visuals.

So, between now and October when these ship, it’s worth auditing the first two sentences of every About page, executive bio, and key product description your company and work includes. Read them out loud. If they don’t hold up without the rest of the page behind them, they need work.

What I’m Sitting With This Week

There’s an odd but very apparent through-line this week that I didn’t fully see until I started writing this: all of it — the vibe coding, the Google Zero panic, the pope, the booing students, the voice glasses — is really about the same thing.

What do humans do now? What is ours to keep?

I don’t think the answer is to refuse the tools. I built a whole app this week using AI that’s genuinely useful. But, I also think we need to be honest about what we might lose along the way.

It’s a lot to think about for sure.

What's your rubber-band golf game building set— the thing you hope AI never touches? Share below and let’s discuss!

Leave a comment

You Beta Believe It is a newsletter for communications professionals about learning AI in public, in real time. If this one made you think or you enjoyed it, please share it!

I have a confession.

These days, at any given moment, I have approximately 15 AI tools running on my computer simultaneously: Wispr Flow, Granola, Fireflies, Notion, Comet, Codex, Claude Chrome extension, Gemini, Perplexity, Midjourney, ChatGPT, Loveable, Replit, Zapier, and Jasper to be exact.

Sure, I could call it research. I mean, this substack won’t write itself … well actually if I deployed an an AI agent, it could, but I digress …

It feels like I’m collecting AI tools the way my 14 year-old daughter collects emotional support water bottles — one for every situation, deeply convinced each one is solving a specific problem, most of them sitting untouched after the first week.

The problem isn’t that there aren’t enough good AI tools, it’s that there are too many, and they’re all slightly different, cost money, and bouncing between them all day is making me a bit of a maniac.

Wait, Harvard Has a Name for This?

Harvard Business Review surveyed 1,488 full-time U.S. workers and found that a chunk of them were experiencing what the researchers called AI Brain Fry — mental fog, headaches, slower decision-making. The kind of feeling where you’ve been “working” all day and have nothing to show for it.

Prevalence varied a lot by role. Only 6% of people in legal reported it. Marketing? 26%. Highest of any function … go figure 😞.

After marketing: people ops, operations, engineering, finance, and IT. So basically everyone except the lawyers, who are dealing with their own separate AI crisis.

The article’s finding that AI tools are intensifying work rather than simplifying it didn’t exactly shock me. Every new subscription comes with hidden overhead. From remembering what the tool does, to exactly when to reach for it, how to prompt it, how to connect or automate it, where your files ended up, and whether you’ve hit the monthly limit. It honestly feels like a part-time job managing the stuff that was supposed to help you do your actual job.

So What Should Actually Be in Your Stack?

One thing I’ve found helpful amid the constant flood of AI tool announcements is thinking about AI stacks in layers instead of individual products. That shift alone made the landscape feel much less overwhelming and helped me stop treating every new launch like an urgent decision.

The basic framework is fairly simple: a handful of core layers, with a small number of tools serving each one. Not the entire AI app store.

The specific tools inside each layer will depend on your organization, budget, compliance requirements, and, realistically, what IT will approve. But this framework helped me recognize where I was doubling up unnecessarily and where I had actual gaps.

The Nice-to-Haves

More advanced PR intelligence platforms, SEO optimization suites, more advanced design and video tools, audience research platforms, niche automations. These can absolutely add value when they solve a defined problem inside an already-functioning process. Buuuuut, they’re also the category most likely to drain your budget while convincing you you’re becoming more productive.

If the answer feels fuzzy, there’s a good chance I’m keeping it out of habit, anxiety, or fear of missing something. That’s been a surprisingly useful gut check for me.

Subscribe now

🧪 Beta Lab: Building Workflows

Goal: Figure out which AI workflows are actually worth building and which ones I just thought sounded good.

Tools: Claude, Perplexity, Notion, Granola/Fireflies, ChatGPT, Zapier, Canva, Muck Rack

The Experiment: I looked at my 15-tool situation and made myself document what I’m actually using, not what I downloaded in a determined New Year, New You frenzy in January.

The Results: These workflows and tools made the cut.

Workflow #1: Press Announcement Plan

Here’s the actual sequence I use now:

• Notion — for building out the plan and housing the structure

  • (🎁 Bonus: Here’s a PR announcement campaign template *here. Just hit copy and then tailor it to your own needs.)

• Claude — for assistance in drafting the release

• Perplexity — for verified research to sharpen media angles

• Muck Rack — for pulling media lists and monitoring coverage

• Canva — for asset development

• Zapier + Airtable —for connecting/automating. So, for example, once a release is approved in Notion, one trigger then prompts a Slack notification to the team, sends a draft to Muck Rack, and saves all assets saved to the shared drive, for example.

(Enterprise teams run this same logic on platforms like Workato or MuleSoft. For smaller businesses, Zapier does the same job for a fraction of the cost and drawn-out IT implementation process.)

Workflow #2: Talking Points and Q&A Development

• ChatGPT — for creating executive and customer personas (custom GPTs)

• Granola or Fireflies — for recording calls/transcriptions of meetings

  • Yes, I see the 🐘 in the room. I just spent this entire post arguing for a leaner stack, and here I am adding tools. I know. But hear me out.

    Recording and transcription tools have also been getting some well-deserved scrutiny lately — consent issues, compliance concerns, and employees finding out they’ve been recorded without knowing. It’s a real conversation and not one to brush past. My take: these tools are genuinely useful for comms work, but only if your legal team has signed off and everyone on the call knows it’s being recorded.

    And, for what it’s worth, I’d put this in the “nice-to-have” category, not the core stack. It’s useful when the workflow calls for it, but not something that needs to run all the time.

• Claude — for feeding the transcription file(s) and any other pertinent documents or website links related to the topic. Then you prompt it to draft 15–20 targeted questions, as well as a one-page document of focused talking points.

• ChatGPT personas — for taking the questions Claude devises and running it through the personas to tailor the talking points for varied stakeholder perspectives.

• Notion — for housing all of it

  It’s also worth mentioning that NotebookLM is a genuinely great AI tool for this type of work. You can upload multiple documents, links, and source materials, and it will analyze across all of them, surfacing themes, questions, and connecting dots in really helpful ways. It’s also able to read links that can be blocked by authentication walls that Claude can’t read.

✏️ Your 20-Minute Homework/Lean Stack Audit

A quick gut check if you want to actually do something about this:

Step 1: List every AI tool you’re currently subscribed to or have installed, including the random ones you read about and got excited to download but haven’t used since the New Year/New You January engery faded.

Step 2: For each one, answer: What workflow does this support? When did I last use it? How much does this subscription cost?

Step 3: Apply this filter: Does it improve at least one of the four of these for me—speed, quality, consistency, and/or measurability?

The strange thing about AI work right now is that most of us aren’t actually suffering from a lack of capability. We’re suffering from too much possibility. Too many tabs open. Too many tools promising transformation. Too many people online acting like if you aren’t testing every new platform within 48 hours, you’re already behind. But most communications work still comes down to the same core things it always has: clear thinking, strong judgment, good instincts, organized systems, and understanding people. AI can absolutely help with that. Sometimes dramatically. But at a certain point, the goal stops being building the biggest stack and starts being building one you can actually think inside of without frying your own brain in the process.

What are your must-have AI tools and the ones you thought you’d use every day but haven’t touched since January? Drop it in the comments.

Leave a comment

You Beta Believe It is a newsletter for communications professionals about learning AI in public, in real time. If this one made you think or you enjoyed it, please share it!

Picture a scenario playing out across communications teams right now: A junior team member sends over a clean, organized, and current media list. Bravo! You then ask how long it took to pull (typically a task that can easily eat up an entire work day). “Twenty minutes,” they say. They had AI pull the contacts and cross-reference them. They just cleaned it up.

Or this: A first draft of a press release lands in your inbox an hour after you handed off the brief to a coordinator, and it’s not bad. Actually, it’s pretty solid. They used AI to draft it, and they shaped it. So, you open it up to start editing, only to realize you’re not really editing so much as reviewing.

Or the market research summary that would have taken a week to compile is in your hands within 24 hours. Synthesized, organized, and ready to present. One person with one AI tool. Done.

And underneath the surface of these moments, simmers the unspoken question humming along: What exactly am I leading now?

I’ve been sitting with it for a while, and I’m guessing a lot of senior communicators are too.

The Headlines Are About Junior Roles, but that is only Part of the Story

PRWeek ran a piece in March (by Chris Daniels, worth tracking down if you have access as it’s paywalled) documenting the emergence of brand new AI job titles across communications and marketing: AI Ethics Officer, AI Solutions Architect, AI Acceleration Lead. New roles, new org chart lines, new budget conversations. At the same time, traditional junior comms roles such as the coordinator, the assistant, the associate are being restructured or simply not backfilled now when someone leaves.

And, while most of the coverage in this piece frames this as an entry-level work issue, the reality is that it isn’t confined to the bottom of the org chart. It’s working its way up, and the question it raises at the senior level is a harder one.

For most of my career, earning praise in communications meant mastering execution. Fast, accurate research. Strong first drafts. Tight media lists. Sharp monitoring. Booking the interview. The ability to synthesize complicated information and turn it into messaging that actually made sense.

That competence was how you built credibility. You moved up because you were fast, reliable, and consistently produced strong work. You earned the senior seat by becoming excellent at execution first, then eventually building and leading the team doing the work you once did yourself.

Now AI can do a version of almost all of it.

Not perfectly. Not without human oversight. But well enough, and fast enough, that execution is no longer the differentiator it once was. Which means the definition of senior-level communications value is changing in real time.

But, here’s what I don’t hear enough people saying directly—it’s not just junior roles that are being compressed, but also the traditional pipeline for developing senior communications talent. The work that once built judgment including the drafting, the research, the monitoring, the grinding through endless media lists until you instinctively understood the beat landscape, is exactly the work AI is starting to absorb.

Which raises a complicated question: if emerging communicators aren’t building judgment through that executional foundation, what does the path to senior leadership actually look like? And on the other end of the org chart, what does the CCO role become when the team you’re leading is increasingly a blend of humans, AI agents, and automated workflows rather than a traditional junior-to-senior hierarchy?

I don’t think anyone has fully figured that out yet. I certainly haven’t.

But, I do think communications leaders need to start wrestling with those questions now, before the structure of the profession gets reshaped for us instead of by us.

What the Ragan Research Is Actually Telling Us

Ragan published their 2026 Communications Benchmark Report in late April, and there’s a stat I keep coming back to: only 5% of communicators said they are reducing headcount because of AI. The report then implies this moment as a “window of opportunity” to redeploy communications teams toward more strategic value. While I love this, I also know exactly what it means— figure this out now, before the window closes.

“More strategic value” can sound abstract, though, until you name it concretely. One place that strategic value is becoming increasingly tangible is earned media. Communications strategist Zen Media’s Sarah Evans recently published a framework reexamining how organizations should think about media coverage in the AI era, and it reframed something important for me: AI systems are now a primary layer through which buyers and audiences discover information, and those systems have a specific (and nontraditional) content preference. They weight transcripts. They favor LinkedIn, YouTube, Substack, and niche trade publications. And, they prioritize recency, with content updated in the last 30 days earning significantly more AI citation eligibility than older material.

What this means for communicators is that earned media now carries two kinds of value: the traditional visibility and reach, and something newer: citation eligibility inside AI-generated responses.

A thoughtful feature in the right trade publication, a well-placed podcast interview, a LinkedIn article from an executive — all of it feeds the information layer AI systems increasingly pull from when someone searches a category, company, or topic.

And, this changes the strategic importance of communications work.

Building media relationships, crafting smart pitches, and securing coverage in indexed, transcribed, citable outlets is not something AI can meaningfully replicate on its own. It requires human credibility, human relationships, and human judgment about what angle will resonate with a specific journalist, at a specific outlet, at a specific moment.

And as AI-generated content floods every other channel, that kind of trusted, attributable visibility only becomes more valuable.

So What Do You Actually Lead Now?

Here’s where I’ve landed, at least for now.

✍🏻️📰You’re the editor. When AI produces a draft, a media list, or a research brief, your job is to interrogate it. Is this accurate? Is it useful? Is it appropriate for this moment, this audience, this relationship? That editorial judgment is yours, built from experience, and it isn’t automatable.

☝🏻️🤖You’re managing AI outputs alongside human ones. This is new. And honestly, it’s weird. Nobody trained us for it. Managing an AI-assisted workflow is a different skill than managing a person.There’s no relationship to develop, no professional growth happening on the other end, no “walk me through your thinking.” There’s just output, and either it’s right or it needs to be fixed. Doing that quality-control function well requires deep subject matter knowledge, which brings us back to experience.

⚖️✊🏻 You’re justifying the function in a new way. At some point, someone — a CFO, a CEO, a board — is going to look at the communications team and ask what it does that AI can’t. Your ability to answer that question clearly and specifically is now a core professional skill. “We provide strategic counsel” is not an answer. The earned media and AI citation argument is an answer. Human relationship networks are an answer. Crisis judgment under pressure is an answer. Get specific, and have it ready.

The Ragan benchmark report identifies integrated communications planning, relationship building, business fluency, and emotional intelligence as the top strategic competencies for communicators right now. But, none of those are AI-native skills. All of them take years to build. This is what we need to focus on rebuilding and reprioritizing within the comms “ladder.”

The Bottom Line: The New Titles Are Telling You Something

Back to those PRWeek job titles: AI Ethics Officer, AI Solutions Architect, AI Acceleration Lead. These aren’t just new roles for new people, they’re signals about where organizational leadership thinks value is being created. Senior communications professionals who can’t engage in the conversation about AI governance, AI integration, and how AI affects organizational reputation and communications strategy will find themselves edged out of the rooms where those decisions get made. You don’t need to become a technical expert (that’s genuinely not the point) but you do need to be able to hold your own in the conversation.

🧪 Beta Lab: Testing Final Layer, an AI Agent for LinkedIn Content

Tool: Final Layer (finallayer.com) — an AI agent for LinkedIn personal branding and content creation. $30/month for the Plus plan.

Goal: After going deep on LinkedIn’s 360Brew algorithm shift last week, I wanted to test whether an AI tool could help me build a consistent LinkedIn presence without the time investment that 3-4 posts per week realistically requires.

The Setup: Final Layer asks you to link your LinkedIn profile and load in your resume, preferred news sources (e.g. New York Times, Wall Street Journal, Fast Company, Wired, Business Insider, Marketing Brew, Axios), target audience (Chief Comms Officers, CEOs, comms professionals), and content pillars (my career and story, communications and PR, leadership and people, AI for communications professionals, and vision and predictions) before it starts generating.

It recommends posting 3-4 times per week to build executive authority on the platform. Once configured, it offers two posting tracks— “Posts” for original content ideas and “Community” for engagement opportunities based on others’ content— with 14 ideas queued at any given time and a refresh option for 14 more (and more …).

The Experiment: The volume of ideas alone has real value. Coming up with 12-16 original LinkedIn post concepts per month on your own is genuinely taxing, and Final Layer removes that friction. It’s also a solid industry news aggregator. I’ve found myself better-informed on relevant trends even when I didn’t end up using the suggested ideas. And, the scheduling feature is practical: you can batch-prep a week of posts in one sitting.

The posts themselves, though, are generic. Even with detailed personalization inputs and heavy editing options, the output consistently feels like a polished version of someone who sounds like me but isn’t quite me. I’ve honestly scrapped most of what it has suggested. And, the Community track hasn’t surfaced content I’d actually want to engage with, though I’m genuinely uncertain whether that’s a calibration issue that improves over time.

The question I keep sitting with is one I didn’t expect: how much do I want AI representing me on a platform that’s fundamentally about professional credibility and human connection? 🤔The efficiency argument is real. The identity question is also real. I haven’t resolved it, and I’m not sure I’m supposed to have resolved it yet.

Verdict: The jury’s out. The trial is ongoing. But that conundrum feels like a preview of a bigger question the whole industry is going to have to answer.

Your 20-Minute Homework

Pick one of these three new AI job titles and spend 20 minutes actually understanding what it means.

AI Ethics Officer: This role owns governance questions — bias in AI outputs, data privacy, transparency in AI use, policy compliance. Communications professionals should care because the reputational risk of getting AI ethics wrong is fundamentally a communications problem.

AI Solutions Architect: This is the person designing how AI tools and workflows integrate across an organization. Understanding their relationship to the communications function, and how your team’s AI tool stack fits into the broader organizational picture, matters if you want a seat in that conversation.

AI Acceleration Lead: This role drives adoption and change management around AI. Communications owns internal communications and culture change. There’s significant overlap here that most comms pros haven’t claimed yet.

You don’t need to become an expert. You need to be able to ask an intelligent question in a meeting. That’s the bar, and it’s a reasonable one to clear this week.

Sources: Ragan 2026 Communications Benchmark Report, Mike Prokopeak, April 30, 2026. PRWeek AI job titles, Chris Daniels, March 2026 (paywalled, worth seeking out if you have access). Sarah Evans / asksarah.ai, CEO Operating Note No. 01, Q2 2026.

What version of “what do I actually lead now?” are you sitting with? I’d genuinely like to know where other senior communicators are landing on this. Drop it in the comments.

Leave a comment

You Beta Believe It is a newsletter for communications professionals about learning AI in public, in real time. If this one made you think or you enjoyed it, please share it!

So, I spend a LOT of time on LinkedIn.

Not because I love it, or because I have some finely tuned personal brand strategy. But because as a communications professional who is actively, publicly building her AI credibility, LinkedIn is one of the primary places I’m doing that work. It matters.

So when I started noticing that my posts weren’t hitting the way they used to, I did what any reasonable person would do—assumed it was my content. Too boring? Wrong time of day? The algorithm hates me specifically?

Turns out it wasn’t any of that (or, at least I hope not about the content and hating me part). It was something much bigger. And, somehow I completely missed it.

Meet 360Brew. The LinkedIn Algorithm Update Nobody Really Told You About.

Quietly throughout 2025, and then with increasing urgency over the past few months, LinkedIn rolled out a complete overhaul of how content is ranked and surfaced on the platform. The new system is called 360Brew, and it didn’t exactly come with a press release. I like to fancy myself as decently on top of current industry news and trends, so I had to go and look at how in the actual hell I completely missed this. Turns out, it rolled out without any fanfare. It was all very covert (which honestly made me feel a bit better about missing it entirely). All of us, collectively, have experienced this sudden shift. Many people reported declines in organic reach of 40 percent to 65 percent in the first months after the update.

Anyway, the CliffsNotes version of what 360Brew is that it’s LinkedIn’s proprietary LLM designed to process the 1.3 billion (!) registered users’ individual text inputs, as well as all additional information, context, posts, interactions, and meaning into one comprehensive system. It then builds a “360-degree semantic authority profile” for every person on the platform and uses that to decide how and to whom to distribute your content.

Whereas LinkedIn’s old algorithm rewarded engagement through likes, comments, hashtags, and posting frequently, the new algorithm is much smarter and more demanding. It rewards alignment and expertise.

But, here’s the rub …

Let’s say you’re a brand marketer. Before, if you posted a clever work-from-home tip and it blew up, great! The algorithm rewarded the engagement and kept pushing your content. Under 360Brew, that post would likely sabotage your distribution and not because it’s bad, but because it’s off-topic. The system simply can’t reconcile “brand marketer” with “here’s my favorite productivity hack,” so it simply deprioritizes you.

This new algorithm wants a congruent story. It reads all of you contextually and it rewards those who have a clear, consistent focus.

For those of us with a genuine niche, this is actually good news. But, for anyone who has used LinkedIn as a repository for varying career updates, disparate contact networks, and and array of diverse posts baked in with the “trending” hashtags, it’s time to recalibrate.

A few other things that changed that are worth knowing:

• Hashtags don’t boost content performance. 360Brew doesn’t factor them into distribution at all. Instead, it identifies recurring themes across your content and uses those to place you. Selective and specific hashtags (2-3 at most) can categorize content for a reader and people can still search for specific hashtags to follow on topics they care about. But, the old approach of applying popular hashtags to your post for increased reach is gone

• It detects and demotes AI-written content. The algorithm actively flags content that reads as AI-generated and deprioritizes it. Authentic voice isn’t just good practice anymore, it’s now an indirect distribution strategy.

• “Post and ghost” is a genuine risk. Posting and then disappearing is one of the fastest ways to tank your content under 360Brew. The system rewards those who show up, engage, and respond after publishing. So, don’t just drop a post and log off. Stick around to engage with other posts and respond to those who comment on yours.

• Company pages to a major hit. 360Brew shifted visibility dramatically from company pages to people with executives and employees now generating four times more engagement than company pages. Companies are now primarily using their CEO and employees to drive organic reach on personal profiles, while company pages provide the paid media and analytics layer. Brands that have been relying on their company page as their primary organic reach engine need to pivot.

• The first two hours matter more than ever. Editing your post within the first two hours after publishing resets the algorithm’s testing window which can completely undercut whatever early momentum you had. Be sure to proofread before you post.

❗But stay with me here, because this is where LinkedIn stops being just a LinkedIn story. I buried the lede a little. Sorry.

What your employees and executives are posting on LinkedIn is now feeding directly into the AI answers your prospects receive when they ask an AI tool who the experts are in your industry. The companies whose teams are consistently publishing expert-level content are quietly building a citation library that AI systems draw from. And, the companies who aren’t are invisible in those same answers.

In short, implementing intentional and updated LinkedIn strategy is much more than merely leveraging a social media channel, it’s a key AI search visibility strategy.

The Employee Advocacy Piece — and Why It’s Now a Comms Business Strategy, Not an HR Program

As I mentioned, 360Brew didn’t just shift visibility from company pages to personal profiles, it created a structural argument for something that comms and marketing teams have been trying to sell internally for years: employee advocacy programs. And now, finally, the business case basically makes itself.

The logic is straightforward. If company pages are getting deprioritized while personal profiles are getting amplified, and you want your brand to have reach and credibility on LinkedIn, you need your CEO, your executives, and your employees posting original content related to their industry (not merely resharing company page content) consistently from their own accounts. Easy, right? 😉

Think of it as your new lead generation engine. When your team consistently posts around two or three defined topic pillars, LinkedIn starts recognizing those individuals — and by extension, your company — as authority in your space. That credibility translates directly to warm pipeline.

So what does a modern LinkedIn employee advocacy program look like?

First, it’s about not coercing, er asking employees to reshare the company page post on Monday morning, especiallyas it will not generate any reach and contribute zero personal authority for the people doing it.

A real program starts with individual expertise, not company content. It defines two or three content pillars that align with business goals and gives employees something worth saying. It provides prompts, support, and editorial guidance without ghostwriting everyone into the same voice. And, it keeps participation voluntary and lightweight because the moment it feels like a mandate, engagement crumbles.

The companies doing this well are seeing their executives and employees become the primary organic distribution engine, while the company page shifts into its supporting role intended for official announcements, recruiting opportunities, paid media, and legitimacy.

For communications professionals: this is our program to own because it requires editorial judgment, message discipline, voice development, and a clear understanding of how credibility is built over time. That is exactly what we do.🫰🏻

✏️Your 360Brew Action Plan

Okay, enough context-setting. Here’s what I’m already doing:

1. Audit your profile first. Your LinkedIn profile is now the lens through which the algorithm evaluates everything you post. Make sure your headline, about section, and featured content clearly reflect the two or three topics you want to be known for.

2. Pick your two or three industry-related topics and commit. Topic drift (posting about too many unrelated things) actively undermines your authority profile under 360Brew. The algorithm can’t position you as an expert in anything if you lack consistent focus.

3. Go deeper, not broader. Under the old system, posting frequently won the game. Now it’s quality and depth over volume. One well-developed post with a clear point of view beats three shallow takes every time. Think: professional insight, specific experience, and direct viewpoint. And, to that point, visual content that illustrates your concept such as a PDF guide, tutorial, or infographic is a great way to promote your expertise and invite user saves (which 360Brew loves).

4. If you’re in comms, build the employee advocacy case now. The shift from company pages to personal profiles is a structural change in how B2B visibility, credibility, and AI search presence are built. Come to that conversation with the numbers: 4x engagement, AI citation data, lead gen implications. That’s how you turn “we should do more LinkedIn stuff” into a funded, staffed program.

5. If you’re building an advocacy program, start with pillars, not policies. Define two or three content themes that align with business goals. Give people something worth saying. Provide prompts and editorial support and keep it voluntary. The moment it feels like a mandate, participation drops off a cliff and you’re back to square one.

🧪 Beta Lab: Lilypath™

Goal: To understand how AI actually interprets my LinkedIn profile and what I can do to show up more authoritatively in AI-generated search results.

Tool: Lilypath™, a patent-pending Authority Intelligence™ platform that analyzes your LinkedIn presence and tells you how AI systems read and rank your professional authority.

There are three package options geared for students ($149), job seekers ($199), and authority builders ($499).

The Experiment: To get started, I uploaded my LinkedIn profile link, a PDF export of my profile (genuinely did not know this was a thing — go to the Resources tab on your profile and hit “Save to PDF” — it exports into a clean, well-formatted document that’s pretty awesome, actually), my resume, and a few other relevant materials. Within minutes, I had a full Lilypath Blueprint report.

The Results: I received an AI Readiness Score of 71 out of 100, landing me in the “Solid Foundation” category which is defined as “decent structure; significant opportunity exists.”

The score is built across five dimensions: Headline effectiveness, About section strength, Experience validation, Skills & Endorsements, and overall profile coherence.

The specific content ideas provided were genuinely useful.

For example, my current LinkedIn headline:

“SVP Corporate Communications | Strategic Communications | Reputation Management | Stakeholder Engagement | Translating Complex Issues into Compelling Narratives That Drive Action and Protect Brand Integrity”

Lilypath recommended replacing it with:

“SVP Corporate Communications | Crisis & Reputation Strategist | Disney, OWN, Marriott | Fortune 500 to Startup Communications Leadership”

The logic: front-load the primary search term, then drop three recognizable brand names for instant credibility. Shorter. Punchier. Easier for both humans and AI to parse quickly.

For the About section, the key feedback was that the opening needs to function as a hook — clearly communicating what you do, for whom, and why — and that the body needs quantified metrics woven in to validate the work.

The blueprint also provided a full network activation and content strategy, as well as a personalized specific notes, frequently asked questions, and an additional appendix of helpful information and resources.

Verdict: 100% worth it. This is just a snippet version of what was a very comprehensive, very informative, and genuinely useful analysis/guide.

If you’ve been wondering how AI systems are actually reading and ranking your professional presence (and after everything above, you should be wondering) Lilypath provides a concrete, specific answer with actual steps for you to act on. It’s not a vague “optimize your profile” checklist. It’s an actual blueprint.

The Bottom Line

This has been a real shift, and not a gentle one. LinkedIn organic reach tanked for a lot of people. Strategies got stale overnight. And the platform offered zero explanation. But, understanding what 360Brew actually wants makes it feel a lot less random and a lot more workable now.

If you want to go deeper on this, I found a few resources genuinely helpful:

⭐ Melanie Goodman, The Link Tank — she specializes in LinkedIn strategy

⭐ This Inc. piece on 360Brew and what’s actually working: inc.com

⭐ Sprout Social’s employee advocacy guide is solid if you’re building a program from scratch.

Your 20-Minute Homework

1. Go download your LinkedIn profile as a PDF right now. Read it like a stranger would. Does it tell a clear story in the first three sentences? If not, that’s your starting point.

2. Pick your two content pillars. Write them down. Then go look at your last ten LinkedIn posts and see how many actually fit. The gap between those two things is your roadmap.

💭 Have you noticed a change in your LinkedIn reach over the past year? I want to know if others are seeing what I’m seeing — drop it in the comments if so!

In honor of Earth Day, I wanted to tackle a subject that’s been hovering in the background of my AI workflows for a while. We’re diving into the world of AI data centers: what they actually are, why their environmental footprint is becoming impossible to ignore, and what we can do about it.

First, AI isn’t going anywhere. And, we can’t holistically deem it villainous and turn away from it because it’s “bad.” Well, at least that’s not what I’m going to do, and I’m guessing most of you won’t either. Some of the best innovations in human history — cars, planes, spacecrafts — contribute to pollution, and we didn’t abandon them. We eventually (albeit imperfectly, inconsistently, and rather ambiguously at times) worked to understand the cost and do better.

AI is the same deal. We need to understand what we’re working with. So I went looking.

An Interesting Parallel to Fast Fashion

Through my work in communications for an industrial hemp fiber startup, I’ve spent a lot of time inside of the textile industry (not the glossy Anna Wintour version, the supply chain version).

I’ve seen the mountains of discarded clothing sitting in open-air dumps in Chile’s Atacama Desert, the kilotons of microplastics leeching into waterways, and the many gallons of water it takes to produce one single shirt. Before that, “fast fashion is bad,” was just a thing I politely agreed with, but now I actually get it.

AI feels similar.

For most of us, using AI feels a lot like shopping at Zara. It’s fast, a bit addictive, and so frictionless that the cost becomes invisible — especially when you're just tapping away on your phone. Meanwhile, far from our screens, the physical world is quietly paying the price in diverted water and strained power grids

What Even Is a Data Center?

AI feels like magic. You type, it responds. Done. 🪄

But behind every prompt you send is a very real, very physical building, somewhere in the world, filled with servers, GPU and CPU processors, cooling systems, batteries, backup power generators, security infrastructures …. aaaaand a host of environmental problems.

⚡ The Electricity Problem

U.S. data centers used about 4.4% of total electricity in 2023. Which sounds manageable until you realize the entire U.S. lighting system — every home, office, and streetlight — uses about 5%.

By 2028, projections land somewhere between 6.7% and 12%, the high end of which is roughly the equivalent to the total residential electricity use of California, Texas, and New York combined.

💧 The Water Problem

This is the one that got me.

In 2023, U.S. data centers consumed an estimated 17 billion gallons of water, which is the entire annual water supply of the city of Houston. Used in one year just to keep servers cool. And what’s worse, a lot of that water doesn’t come back. It evaporates. It’s just … gone.

By 2028, projections for large-scale facilities alone hit up to 33 billion gallons annually.

To bring it down to a more personal level, scientists at UC Riverside estimate that a 100-word AI prompt uses about one standard water bottle’s worth of water, and there are hundreds of millions of prompts sent every single day.

🌫️ The Carbon Emissions Problem

Data centers run on energy, and much of that energy comes from fossil fuels. One estimate I read puts the increase in emissions tied to data centers’ demands at around 220 million tons of carbon. For context: driving one gas car 5,000 miles produces about one ton of CO₂.

Also worth noting: those diesel backup generators that kick in when the grid strains? Not exactly clean energy heroes.

And We're Just Getting Started

We’re currently at about 11,000 data centers globally. At the current pace of AI growth, data center capacity will need to grow by 130% by 2030.

For a sense of how seriously the industry is taking this, U.S. spending on data center construction has now officially surpassed spending on traditional office buildings for the first time in history. By the end of 2025, monthly data center construction spending reached over $3.6 billion.

And nobody can tell you exactly how many data centers we need, because nobody fully knows how much AI we’re going to use or whether efficiency gains will outpace demand. What we do know is that right now, we’re building as fast as we possibly can and just hoping the math works out. 😰

Ok But What Can We Actually Do?

This is a systemic issue that requires policy, regulation, and corporate accountability to meaningfully move the needle. Individual action isn’t going to fix this kind of structural problem.

But informed users turn into informed employees, and those people ask better questions in the rooms where decisions actually get made.

For Individuals🧍🏻‍♀️

✅ Use the lighter model

Most of us pick one AI tool, use it for everything, and never think about it again. But every major platform has model options and the lighter, faster versions use a fraction of the energy of the big flagship ones.

Inside Claude: there’s a model selector at the bottom of the chat window. Haiku is the lightest and great for quick summaries, rewrites, and short Q&As. Sonnet is the middle ground, Opus is for genuinely heavy lifting.

Inside ChatGPT: GPT-4o mini is the lighter option. There’s also an “Auto” mode that picks the right model size based on your request.

The rule of thumb: if you’re about to hit send on something routine such as summarizing a document, fixing a sentence, generating a quick list, try the lighter model first. You’ll probably get 95% of the quality at a fraction of the footprint.

✅ Match the tool to the job

Think of your AI tools less like a Swiss Army knife and more like a small team with different specialties:

• Claude → deeper writing, editing, long-form thinking

• ChatGPT → drafts, ideation, quick iterations

• Gemini → research inside Google’s ecosystem

• Perplexity → fast, cited lookups

Better tool selection = fewer retries = less wasted energy

✅ Prompt smarter

Every additional prompt costs something. Better prompts with clear, specific, with context upfront reduce trial and error, cut down on follow-up questions, and shorten conversation length. Getting to the point the first time is both a productivity win and, it turns out, a small environmental one.

✅ Ask questions at work

If your company is rolling out AI aggressively, it’s reasonable to ask what that actually costs environmentally. You don’t need to be the expert, you just need to be the person willing to raise the question.

A Note on “Choosing the Greenest Provider”

You may have heard the advice to choose AI providers with strong renewable energy commitments. It’s good advice in theory. In practice, it’s complicated.

Here’s a breakdown on the major players:

Google (Gemini) is the most vocal. They’ve matched 100% of their global electricity use with renewable energy purchases since 2017. But their own 2025 environmental report acknowledged that total emissions increased 11% year-on-year due to data center growth. Their emissions are up 51% from their 2019 baseline.

Anthropic (Claude) has not publicly committed to specific climate goals, has not reported carbon emissions figures, and has no documented reduction targets on record. They run on Amazon Web Services, which reached 100% renewable energy in 2023, so there’s an indirect green claim via the cloud provider. But Anthropic itself has published essentially nothing on its own environmental footprint.

OpenAI (ChatGPT) scores a 23 out of 100 on DitchCarbon’s sustainability rating reflecting a near-total absence of disclosed environmental data. No annual sustainability report. No third-party audited emissions figures. Their primary defense is that they run on Microsoft Azure, which has its own green commitments.

The uncomfortable truth is that right now, choosing the “greenest” AI provider is a bit like trying to pick the most ethical fast fashion brand. The options are limited, the data is murky, and the most important thing might just be using less of all of it and more thoughtfully.

The Part That’s Relevant to Your Job

Companies are racing to adopt AI and many have made public climate commitments. Those two things are increasingly at odds. Simultaneously, employees are noticing and the tolerance for vague sustainability language or greenwashing is dropping fast.

According to S&P Global’s 2025 Corporate Sustainability Assessment only 36% of companies have a dedicated AI governance policy at all, and fewer than 1 in 5 can actually quantify the impact of their AI use on sustainability goals. It’s not a great foundation for credible communication, but it’s exactly where comms professionals can add real value.

What employees actually want to know

They’re not asking for perfection. They’re asking for clarity on these main questions:

• Do we understand what AI is costing us environmentally?

• Is leadership making decisions about AI adoption with sustainability factored in, or is it just speed and efficiency?

• What can I do, in my own role, that makes a difference?

• Are we being honest about the tension between our AI growth goals and our climate commitments?

The worst outcome isn’t that your organization doesn’t have all the answers yet, it’s that employees sense the answers exist but aren’t being shared.

What this looks liks is clear internal communication that involves stating the tension out loud, sharing what you know even if it’s incomplete, and giving people something practical to do or contribute to as you all figure it out together.

The companies that handle this well won’t be the ones with perfect numbers, they’ll be the ones telling the truth while they figure it out.

The Bottom Line

The infrastructure behind AI is heavy, physical, and resource-intensive in ways most of us—myself included—rarely stop to consider. The water evaporates. The fossil fuels burn. Data centers are rising faster than renewable energy can keep up. I’m not sharing this to catastrophize; I simply looked under the hood, and this is what I found.

But, none of thsis means we log off and call it a day. It just means that we use these tools with a little more awareness, a little more deliberately, and maybe a little more proportionately, and we stay in the conversation about what better looks like.

Knowledge is power, and the people who understand the cost are the ones who eventually will help to change it.

Happy Earth Day. 💚

📬 If this made you think differently, send it to someone who uses AI all day without thinking about what’s behind it.

Did this change how you’ll use these tools?

There’s so much happening in AI right now that it’s hard to separate signal from noise, especially if you’re not living in it every day. But if you work in communications, some of these developments aren’t just interesting, they’re dramatically shifting how we do our jobs.

This week, I wanted to share a handful of the top AI-related new stories I’m paying attention to right now and what they mean for comms teams.

And, disclaimer: I think after this week we’re allll just a little afraid of Little Ole’ AI 😬

1. The New Yorker’s damning profile of OpenAI CEO Sam Altman

The New Yorker recently published a long, deeply critical profile of OpenAI CEO Sam Altman, the result of over 18 months of reporting, more than 100 interviews, and a trove of never-before-seen company internal documents. Written by Pulitzer Prize-winning investigative journalist Ronan Farrow and political reporter Andrew Marantz, it is truly fascinating, albeit verrrrrrry long. (Sidenote: the accompanying AI-generated animated image they used for the online article of a Medusa-esque, multi-headed Sam Altman is incredibly, and intentionally, creepy😱)

The central tension of the piece is between Altman’s public persona and his internal leadership style—and ultimately whether he’s the right person to steward a technology this powerful. The part that really stuck with me: a board member describes Altman as possessing two rare and deeply conflicting traits—“a desperate need to be liked in every interaction, paired with an almost sociopathic lack of concern for the consequences of deceiving others.” He’s consistently characterized as being “unconstrained by truth.”

To say this story has dominated political and tech news coverage lately is an understatement. If you want to go deep on it, I recommend listening to this NYT “Hard Fork” podcast episode.

👀 Why it matters for comms:

This profile is already shaping how policymakers, journalists, and industry leaders talk about AI governance and trust. If your company relies heavily on a single AI provider like OpenAI, you are tethered to that leadership team’s reputation. A bombshell investigation doesn’t just affect them, it affects every company using their technology, and how stakeholders interpret your choices.

The term “responsible AI” is getting harder to say with a straight face. Audiences are becoming skeptical of it, and rightfully so.

👉🏻 What to consider:

• Create an internal AI governance one-pager for executive and board use that clearly documents who owns what in AI, how decisions get made, and how concerns are escalated (add this to your overall company AI policy).

• If your external-facing language (think website, FAQs) uses words such as “safe,” “trusted,” or “responsible” AI, now would be a good time to make sure those claims can hold up to scrutiny.

2. OpenAI Bought a Talk Show

OpenAI acquired TBPN (Technology Business Programming Network), a notably-popular three-hour daily tech and AI talk show on YouTube and X, for a reported “low hundreds of millions” sum on April 2.

At first glance, this looks like a very expensive, very weird flex. But look closer.

OpenAI basically decided that the traditional comms playbook doesn’t work for them anymore. Since they’re not merely selling toothpaste, rather they’re introducing a world-changing technology that most people don’t fully understand, regulators are circling, and narratives shift daily, they determined that they’d prefer moving from defense mode (reacting to media coverage) to offense mode (buying a seat at the table where stories get written).

TBPN’s value isn’t its modest revenue, it’s in their trusted audience, their direct access to key builders and decision-makers, and their unique format where conversations feel unscripted. That’s incredibly hard to build from scratch, and much easier to acquire.

Multiple reports have explicitly framed this as a very smart (cost effective) marketing initiative, rather than a media investment. Instead of buying ads, they own the platform. Instead of pitching reporters, they host the conversation. Instead of fighting narratives, they shape them early.

(Worth a listen: separate interviews with OpenAI’s Chief Policy Officer Chris Lehane and TBPN president Dylan Abruscato on Puck’s “Grill Room” podcast)

👀 Why it matters for comms:

This deal is a signal. Media, marketing, and communications are converging, and the old assumption that talent, media, and distribution are separate assets is breaking down.

And, top-tier companies no longer want to wait for a journalist to “get” their story, they want to own the platform that tells it.

👉🏻 What to consider:

You don’t need to buy a talk show. But you can stop thinking like a publicist and start thinking like a publisher. Build repeatable owned assets like a newsletter, a briefing series, a podcast, a consistent presence where your audience already is. Keep doing media relations, but stop treating it like the engine. The comms pros who are going to matter most in the next few years are the ones who can build and hold an audience for their business/brand independently.

Subscribe now

3. The Disney and Snap Layoffs Hit Differently

Two big layoff announcements this week that deserve your attention.

Disney confirmed it’s eliminating approximately 1,000 jobs, largely within its marketing and brand organization, including entire home entertainment publicity and EPK teams, and senior digital marketing and publicity leaders.

Snap also announced it’s cutting roughly 16% of its global workforce (around 1,000 roles plus 300 open positions) to save more than $500M annually and, in their words, “reposition the company around AI-driven efficiency.”

In both cases, leadership is explicitly linking the moves to cost discipline and the belief that AI enables smaller teams to move faster and do more. This isn’t an isolated story. From Amazon to Pinterest, Meta to Nike, Oracle to Angi, the layoff trend over the last few months, often explicitly tied to AI, automation, or “refocusing on AI-driven priorities,” is real and accelerating.

👀 Why it matters for comms:

Disney cut whole publicity teams, including executive and VP-level roles— functions that historically sit close to the core revenue engine of the studio.

Efficiency is the new priority. In the current economic climate, leadership is less interested in “creative brand sentiment” and more focused on communications that drive immediate, measurable business outcomes. If a role is focused purely on traditional PR or top-of-funnel awareness, it can be viewed as “discretionary” rather than “essential.”

We can expect more and more CEOs to argue that they can hit the same or greater output with fewer marketers and comms pros because of AI‑enabled content production, targeting, and analytics.

👉🏻 What to consider:

• If your role is currently focused on top-of-funnel awareness or traditional PR outputs, that’s a positioning problem worth addressing before someone else addresses it for you. Think about reframing your position from “storytelling and promotion” to business infrastructure: reputation risk management, investor and policy narrative, pricing power, and deal support. Build dashboards or case studies that show direct impact on business outcomes (e.g., launch success, cost of crisis averted, stakeholder behavior change).

• Lean into AI fluency, but frame it as force‑multiplier, not replacement. Large-scale enterprises are realizing they can maintain the same output with fewer people by using AI to handle "lower-level" comms tasks like initial drafting, media list curation, and basic sentiment monitoring. The professionals who survive these cuts are typically those who have mastered the orchestration of these tools rather than those whose jobs are being automated by them. You can also emphasize where human judgment is non‑negotiable: crisis calls, ethics issues, regulatory interpretation, stakeholder coalitions, and complex reputation tradeoffs for example.

4. Anthropic’s Mythos Cybersecurity Situation

Anthropic recently released (but significantly restricted) a model called Mythos because of how powerful and potentially dangerous it is. It can identify and exploit software vulnerabilities autonomously, in a matter of hours rather than weeks. Regulators and central banks are already raising alarms about its potential to supercharge cyberattacks, particularly in fragile systems like banking infrastructure. Governments and companies are scrambling to test their defenses before broader availability.

This is the first widely acknowledged example of AI that can materially accelerate real-world harm at scale (go on and cue Darth Vader, I mean geez).

👀 Why it matters for comms:

The communications challenge here isn’t just about Anthropic, it’s about what any company deploying AI now has to contend with: how do you communicate uncertainty without losing trust?

Fear spreads faster than facts. Mythos coverage is being driven by “unprecedented risk,” “national security concern,” and “cyberattack escalation.” Comms has to neither dismiss those concerns (which kills credibility), nor amplify panic (which creates backlash).

The answer, as uncomfortable as it is, is to acknowledge ambiguity, show active risk management, and communicate process rather than conclusions. If you’re in healthcare, financial services, public policy, or any sector where confidential data is fundamental to the business, this news announcement and the ability it has to reshape cybersecurity is a particularly overwhelming moment to navigate. There is zero playbook for anyone here, we’re all learning as we go.

👉🏻 What to consider:

• Acknowledge ambiguity, show active risk management, communicate process, not just conclusions.

• Make sure you are consistently working on and optimizing your AI crisis communications planning to match what’s happening, as the AI capabilities are moving fast.

The Bottom Line

Every one of these stories — Altman’s character and what it means for trust in AI, OpenAI buying its own narrative and marketing infrastructure, Disney and Snap cutting comms functions, Anthropic releasing a model that poses a serious threat to company security — is really about the same thing. The playbook that comms pros have relied on for the last twenty years is being rewritten by the same technology we’re all still figuring out how to use 🤖.

Which is either a terrifying or a great time to be paying attention.

I’m choosing the second one. Mostly. 🥴

Based on this round of news, I have two questions you can choose from here:

1. Which of these stories hit closest to home for you — and are you treating it like a warning or a to-do list?

2. If someone asked you tomorrow to prove your role couldn't be automated, what would you say?

   Please share in the comments and let’s discuss!

Leave a comment

You Beta Believe It is a newsletter for communications industry professionals about learning AI tools in public, in real time. If you found this useful, the best thing you can do is share it with someone who needs it. I’d be so grateful! 🙏🏻

Share

Let’s just say it: if your nervous system feels like it’s been running a marathon it never signed up for, you’re not imagining things. We are living in a full-time, no-days-off crisis news cycle, and the whiplash is real.

Tariffs, layoffs, AI panic, war, ICE raids, and a general sense that someone left the door open and all the chaos just walked in. Our collective cortisol levels are through the actual roof.

When the world outside is spinning, the world inside of your organization starts spinning, too. Emotions are heightened, employees are distracted, leadership is reactive, and the line between “external noise” and “our actual crisis” gets blurry fast. This is precisely the environment communications professionals are walking into every single day right now. Add in a dash of inflation, a pinch of government shutdowns, and a scoop of geopolitical turmoil, companies are getting squeezed from every side while frustrated consumers take to social media to make their feelings very, very known. The result is a perfect storm that communications teams didn't exactly train for: external crises colliding with internal anxiety, in real time, in public, with everyone watching. 😵‍💫

And … while this is happening, AI is gathering speed because the internet doesn’t wait. When a crisis breaks (big or small), consumers are already posting, venting, filming, and sharing in real time.Thousands of posts, news clips, and social reactions get synthesized and amplified before you've finished your first internal briefing.

Three AI-Related Crisis Scenarios

So, deep breaths here folks because I promise to get to the tools/solutions/AI upside of all of this (somewhat 🥴) if you just hang in there with me.

Let’s first look at three specific and pervasive scenarios crisis comms teams need to be prepared for:

⚠️ 1. Synthetic Crises: Deepfakes and Voice Clones

If you think deepfakes are still a “someday problem,” the numbers are startling. According to elite cybersecurity firm DeepStrike, the volume of deepfakes grew from roughly 500,000 online in 2023 to an estimated 8 million by 2025, with annual growth nearing 900%. And, get this, scammers need as little as three seconds of audio to create a voice clone with an 85% match to the original speaker.

The scenario crisis comms teams need to prepare for are deepfake videos of your CEO or C-Suite team making harmful claims or remarks that people believe or act upon. Human detection of high-quality deepfake videos is only 24.5% accurate SQ Magazine, which means your audience almost certainly cannot spot it on their own.

The communication challenge is no longer just “respond fast.” It’s “prove you’re real.” Many companies are now proactively adopting digital content credentials, specifically C2PA standards, for media that verifies the authenticity of your actual assets before a crisis ever hit because once a deepfake is out there, you’re not just fighting misinformation, you’re fighting the internet’s willingness to believe it.

⚠️ 2. Algorithmic Firestorms and GEO Damage

Traditional SEO was already a full-time job, but now we have GEO (Generative Engine Optimization) to understand and optimize for. When consumers search for information today, they are increasingly getting AI-generated summaries instead of a list of links, and those summaries are built on whatever the internet was saying about you when a crisis hits.

The Marriott-Sonder situation offers a masterclass in just how fast a situation can go sideways. If you’re not familiar, in August 2024, Marriott signed a 20-year deal to integrate Sonder’s 9,000+ rental units into its Bonvoy platform. By November 2025, Marriott terminated the deal citing Sonder’s default and within 24 hours of the termination, Sonder filed for Chapter 7 bankruptcy and began immediately liquidating its U.S. assets. Marriott released a statement on a Sunday morning promising to “minimize disruption” and “support guests.” Unfortunately, the operational reality was thousands of guests around the globe were being evicted with a mere 15 minutes notice, some in the middle of winter storms.

Social media, predictably, exploded. Reddit, Threads, and LinkedIn filled with firsthand accounts of people stranded and abandoned. And because Marriott’s polished PR statement didn’t match what was actually happening on the ground, AI search engines had two very different stories to work with. They indexed both, and then synthesized them into summaries that characterized Marriott as “abandoning guests” regardless of the official statement. Marriott’s legal response, which alleged Sonder used “guest safety as a bargaining chip,” only added more conflict data to the pile.

The result? Months later, ask an AI about the Marriott-Sonder partnership and the top-line summary is still built around words like “evictions,” “stranded guests,” and “liquidated.” The narrative wasn’t lost in a news cycle, it’s now locked into an always-on algorithm.

The lesson for communications teams is this: you cannot manage an algorithmic firestorm with a carefully-calculated press release or statement. You need clarity quickly, and you need to assert what you know, what you don’t, and what you’re doing next. (Thank you - more on her crisis comms wisdom below)

⚠️ 3. Hallucinations in Regulated Industries

In the now-landmark case that I’ve cited before, Air Canada learned this particular brand of AI crisis situation the hard way when its customer service chatbot promised a passenger a bereavement discount that directly contradicted the airline’s actual policy. A court ruled that Air Canada was on the hook for it anyway because the bot said it.

That ruling cracked something open because it established that a company cannot point to its AI and shrug. If your technology makes a promise, gives advice, or states a policy, you may be legally responsible for what it says, even if what it says is completely fabricated.

In regulated industries, banking, healthcare, legal services, this is not a hypothetical risk, it’s is an active liability sitting in your customer-facing stack right now. A compliance AI that fabricates a sanctions violation against a legitimate client, or a medical chatbot that fails to recognize a metaphorical expression of distress and responds with something dangerous, are no longer edge cases.

The communication challenge here requires radical transparency. Audiences in 2026 will often forgive a human mistake, but they are far less forgiving of a technical cover-up. When an AI error causes harm, the response cannot be defensive or legalistic. It has to be honest, fast, and accompanied by a clear explanation of what went wrong and what is changing. The alternative, letting AI search engines index your silence and your legal filings (see crisis scenario 2 above), is how you end up defined by the worst version of the story.

✨ Brand Trust is KEY

I want to credit , host of The PR Breakdown podcast and Substack (if you’re in comms, she’s a must-follow), for getting me thinking about all of this. Her work, particularly in her April 1 podcast “AI Already Wrote Your Crisis Story,” emphasizes a tenet I’ve held onto since the earliest days of my career:

Your brand is arguably your most valuable asset. Not your product. Not your technology. Not your market share. Your brand. And the currency your brand runs on, with every employee, customer, investor, journalist, and stakeholder you have, is TRUST. It always has been. But right now, in this particular moment, with everything we’ve just walked through, trust has never mattered more.

Think about what trust actually has to do for you when things go wrong.

• It has to convince people to believe you when you admit you’ve made an error.

• It has to give you the credibility to stand up and say “that video is a deepfake” and have people actually take your word for it.

• It has to carry you through the moment your chatbot says something indefensible and you have to look your customers in the eye anyway.

Trust is not a soft, feel-good concept. It is load-bearing infrastructure. And if you haven’t been building it, protecting it, and treating it like the strategic asset it is, you are going to find out the hard way that there is nothing underneath you when the floor gives out.

Trust is the foundation of what McPherson urges communicators to understand about crisis response in the age of AI. It’s not about which platform to monitor, or how fast to post, but whether you have earned enough trust, before the crisis, to matter during it. Because, as she points out, the media landscape has dramatically shifted twice (with social media first, then AI). She says, “you are no longer the first draft of that story. If you wait for even a short time, the first version of the story gets drafted without you. It’s being summarized at scale.” Which means you don’t get to wait for perfect information anymore. You don’t need every answer. What you need is clarity, and the trust you’ve already built so that when you show up, people listen.

🛠️ AI Crisis Comms Toolkit

Seeing It Coming Before It Arrives

This is where AI has changed everything. Remember when “monitoring” meant someone googling your company name every morning? Those days are embarrassingly over. 👵🏻

• Platforms such as Signal AI and Noggin now analyze sentiment and patterns across news and social media to flag threats before they make headlines. Think of them as the smoke detector, not the fire extinguisher.

• Some organizations are going further, using AI agents to scan anonymized internal communication patterns — think unusual spikes in Slack activity or shifts in employee sentiment — as early indicators of a brewing HR or compliance issue. Tools like Deel HR use AI classifiers to flag localized legal risks in real time, catching workforce compliance issues before they become lawsuits.

Media Monitoring Has Had a Full Glow-Up

We’re not just talking keyword alerts anymore, we’re in the era of what the industry is calling “Narrative Intelligence” (the ability to understand, analyze, and interpret the stories, storylines, and information networks that shape public perception and behavior).

• CisionOne has a “React Score” that gives enterprise teams a real-time read on how their messaging is actually landing across broadcast, print, and social.

• Talkwalker remains a go-to for social-specific crisis alerts.

• TVEyes gives you real-time searchable transcripts of global broadcast and podcast content because your crisis can break on X or cable news at 6pm.

War-Gaming Before the War

This is the use case I think is most underutilized right now.

• CrisisCalm lets you input a specific scenario — data breach, natural disaster, executive misconduct — and generates a full crisis playbook in minutes, including stakeholder-specific messaging and media Q&As.

• Conducttr runs immersive simulations where your team literally “plays through” a crisis across simulated social and news feeds. Muscle memory matters. You don’t want the first time your team has navigated a deepfake scenario to be an actual deepfake scenario.

• AI can also run a “Red Team” (adversarial response) on your own response — feed your preferred AI your draft statement and ask how a hostile Reddit thread would tear it apart. You’d rather find out in rehearsal.

Your Secret Weapon: Your Own Data

Your organization (likely) has years of institutional knowledge sitting in old decks, stale email threads, and someone’s Google Drive folder from 2019. The teams winning right now are building what some are calling a “Data Vault,” or an internal AI library trained on their own proprietary information. Think past crisis playbooks, brand voice guidelines, stakeholder maps, what worked, what didn’t, lessons learned the hard way … alllll of it.

So, when a crisis breaks at 11pm on a Sunday, your AI isn’t starting from scratch. It’s drawing from everything your organization has already lived through. And because the AI is pulling from your verified internal sources rather than the open internet, you also dramatically reduce the risk of hallucinations. (This technique is called “RAG, or Retrieval-Augmented Generation,” if you want to sound very smart at your next staff meeting.🤓)

Platforms like Domo can help democratize this even further, giving legal, operations, and communications teams a comprehensive, tailored view of the same data during a high-pressure event so that everyone is working from the same reality, not three different versions of it.

The Framework Holding It All Together: The OODA Loop

Originally developed by military strategist Colonel John Boyd for fighter pilots, the OODA Loop is increasingly being integrated into AI-powered crisis response — because it’s built for exactly the kind of high-stakes, fast-moving situation we’re describing.

It goes like this:

• Observe — Gather the signal. AI filters out the noise and flags the statistical anomalies worth your attention.

• Orient — Make sense of it. This is where your Data Vault earns its keep — comparing the current situation to past crises to read the trajectory.

• Decide — Choose your move. Proactive statement? Direct journalist outreach? Silence? (Silence is a decision, and usually a bad one).

• Act — Execute. Generative AI versions your core message instantly into a press release, an internal memo, and social responses all in one pass. But, a human needs to give it the final touch here. No AI use with the human element!

Then the loop starts over, because you immediately go back to observe to see how the world responded to what you just did.

Before AI, one OODA loop in a crisis could take 24 hours. In 2026, teams running the right tools are cycling through it in minutes. And here’s the cold truth: if the other side — whether that’s activists, bad actors, or just a viral tweet — can run their loop faster than you can, they control the narrative, not you.

The Bottom Line

Crisis communications has always required fast thinking, clear heads, and the ability to speak plainly when everything around you is on fire. That hasn’t changed.

What has changed is the speed, the scale, and the number of ways things can go sideways before you’ve even been notified there’s a problem. A deepfake doesn’t wait for business hours. An algorithmic firestorm doesn’t pause while you draft your holding statement. AI is summarizing the story of your brand right now, today, from everything that’s ever been written about you and you had zero input into that summary.

So yes, learn the tools. Build the Data Vault. Run the simulations. Optimize for GEO. Do all of it. But start with trust. Build it before you need it. Protect it like the asset it is. Because when the sh*t hits the fan— and at some point, for every organization, it does — trust is the only thing that gives people a reason to hear your side of the story.

✏️ Your 20-Minute Homework

Pick one. Just one. You don’t need to overhaul your entire communications infrastructure this week.

• If you’re in-house: Google your company name in Perplexity, ChatGPT, or Gemini right now (or, better yet, all three). Read what comes back. That AI-generated summary is what your stakeholders are increasingly seeing first. Is it accurate? Is it fair? Does it reflect the brand you’ve been building? If the answer to any of those is “not really,” you just found your starting point.

• If you’re an agency or consultant: Ask your next client when they last ran a crisis simulation. If the answer is “never” or “I think we did one in 2019,” offer to run a basic one with them. You don’t need Conducttr to do it. You just need a scenario, a timer, and a willingness to find the gaps before a real crisis does.

• If you’re a solo communicator or freelancer: Start your own version of a Data Vault. It doesn’t have to be fancy. A Google Doc with your best crisis examples, your go-to messaging frameworks, and the lessons you’ve learned the hard way is a legitimate starting point. Future you will be very grateful.

  Thanks for reading You Beta Believe It ! This post is public so feel free to share it.

  Share

Not long ago, knowing how to write a decent AI prompt felt like a superpower. I was the person in the room who actually knew how to talk to ChatGPT. I had opinions about Claude vs. Gemini. Heck, I was building a whole Substack around it.

And while AI agents have technically been around for a bit, something has significantly shifted with agentic AI in the last few months. The tools to build them got accessible (as in: you don’t need to code). The use cases got real (as in: someone’s agent is clearing their inbox while they sleep—and it’s not just a tech bro in a hoodie). And, the conversation went from niche to unavoidable (as in: boardroom, break room, your mom’s Facebook feed).

The Elephant in the Room

First, let me just address the elephant in the room: I have not built an AI agent yet.

I’ve gone deeeeeeep down the rabbit hole — reading, listening, bookmarking, spiraling. I’ve had approximately 47 browser tabs open on this topic for the past two weeks. I’ve taken a borderline concerning number of notes. Conceptually? I get it. I’m actually proud of how much I do understand. Practically, Have I built one? No … because, I’m still decently intimidated.🫣

So, this week’s post isn’t a how-to. Think of this one as more of me dragging you along while I figure it out—which, if you’ve been here a minute, is very on brand.

So, what actually changed?

The chatbot era had a rhythm. You opened ChatGPT, Claude, or Gemini. You asked a question. It answered. Maybe you refined it, copied it, edited it, used it. It was reactive, contained, and manageable.

An AI agent is a different animal entirely.

The simplest definition I’ve landed on: An AI agent is a system that can reason, plan, and take actions on its own based on a goal you give it. While a chatbot answers a question (brain), an agent autonomously manages it (brain + action).

Think of it as: Brain + Memory + Tools
🧠 Brain: the AI model doing the thinking
💾 Memory: what it knows about you, your work, your preferences
🔨 Tools: what it can actually do (search, email, read docs, update files, make decisions)

But then I had to question what makes an AI agent different from an AI workflow (remember my custom GPT/Zapier integrations for my daily news briefing?), and the distinction that finally clicked for me is that an AI workflow is rigid. It’s if-this-then-that. You build the path, it walks it. If something unexpected happens? It breaks. Or defaults. It’s useful, but not smart. An AI agent, on the other hand, doesn’t just follow a path, it reasons through it. If it hits a roadblock, it doesn’t just stop. It figures out what to do next (within the guardrails you’ve set).

That’s the leap. And that’s what makes this feel like a completely different era.

Why now?

The pieces finally clicked into place.

The models got smart enough. The tools got usable enough. And someone figured out how to connect them and offer it all up in a way to people that doesn’t require a computer science degree.

Industry leaders say that 2026 is the year of AI agents, and by all measures, that’s not just PR hype. But, while there is a lot of talk around job loss due to AI (er, AI washing), agents aren’t replacing entire jobs, at least not yet. They’re replacing parts of jobs — the parts that are repetitive, research-heavy, time-consuming, and low on judgment. What this looks like for comms professionals is that teams are converging around five core types of agents:

* Brand governance
* Research and intelligence
* Content and messaging
* Response and triage
* Strategy support

I shared a few of the regular tasks that comms professionals have with Claude and asked what type of AI agents would be most helpful. Claude shared the concept of building a narrative intelligence AI agent to pull news, social chatter, analyst notes, and competitor moves; then cluster the themes, flag potential risks, and surface opportunities for proactive outreach and thought leadership. Ok yes, I’m in! (well, not really because I haven’t built it.)

A Quick Flashback 💭 (because I can’t NOT share this)

I had a full-body memory while thinking about a narrative intelligence AI agent of my many early mornings spent at the Columbus Circle Equinox gym, just a few blocks from my apartment on 63rd Street.

My publisher, the legendary Bob Miller, had a habit of popping his head in first thing to quiz me on the news of the day. And I was never, not once, going to be caught flat-footed.

That whole routine feels almost charmingly analog now.The goal was exactly the same. The method just involved a lot more cardio. 😂

The Real Takeaway
Most communications professionals are still doing much of this manually.
An AI agent handling the assembly work so that you can focus on judgment is not a small shift.

Subscribe now

So why am I intimidated if “anyone can do this”?

This is the part I’ve been sitting with. Because on paper, everything I’ve read says: You don’t need to code. You don’t need a technical background. These tools are built for people like us. And I actually believe it.

So why haven’t I done it yet? 🤔

I don’t think the barrier is technical. I think it’s cognitive. Building an agent isn’t just learning a tool. It’s learning how to think differently about your own work.

You have to:

• break a task down into its component parts

• decide what matters vs. what doesn’t

• define judgment vs. execution

• trust something else to handle pieces of it

It’s one thing to ask ChatGPT for help and review the output, and it’s another thing entirely to say, “Go handle this,” and step back. That requires a level of trust, and specificity, that most of us (👋) haven’t had to practice yet.

So yes, the tools are more accessible than ever, but the mindset shift will require some work.

Where I am right now

I’m in the research and tool-evaluation phase, which is a fancy way of saying I’m still figuring out where to start.

I’m looking at tools like Relay and Gumloop for a more simplified and all-encompassing approach to agent building. I’ve also been digging into OpenClaw, which is a genuinely wild open-source personal agent that lives on your computer and you control from your phone via WhatsApp or Telegram. Love their tagline, by the way: “A space lobster AI with a soul.” 🦞

I’ve been paying close attention to the thought leaders in this space. For example, Allie K. Miller (AI with Allie), who by the way, if you don’t follow already you absolutely must, shares a free template called the “Agent Protocol“ that provides a detailed framework for you to follow as you build an agent.

The Bottom Line

Unfortunately, I don’t have a tidy conclusion this week because I’m not on the other side of it yet. But, the gap between where I was just six months ago to where this technology is right now feels significant in a way I can’t ignore. While the prompt-and-chat era isn’t over, it’s no longer the frontier.

The frontier is agents. And I’d rather be someone who is decently intimidated but kept going than someone who waited until it all felt comfortable. Scrappy action for the win!

✏️ Your 20-Minute homework

• Pick one task you do every single week that:
  - takes more than 30 minutes
  - feels repetitive
  - requires more assembly than actual thinking

• Write it down:
  “I wish something could just handle ______ for me.”

That’s probably your first AI agent.
No building required yet. Just noticing. We’ll figure out the rest together. 🤓

If you try this, hit reply or drop it in the comments — what’s the first task you’d hand off to an agent? ** I’m collecting real-world examples and may turn a few into actual builds in an upcoming post.

Leave a comment

One of the clearest warnings about AI risk didn’t come from a dystopian headline or a congressional hearing. It came from a chatbot.

In 2024, Air Canada was taken to a tribunal after its website chatbot gave a customer wrong information about bereavement-fare refunds. The company’s defense was essentially: don’t blame us, blame the bot. The tribunal was not impressed. If it appeared on Air Canada’s website, Air Canada owned it.

And, this is why AI matters for communicators: it can do a lot, but it can't take the blame. That part is still yours.

Why this matters now

Inside most organizations, AI use is already outpacing policy. People are drafting with it, summarizing with it, researching with it, and (let’s be honest) pasting things into it that would make their legal team run for the hills. Confidential docs. Proprietary strategies. Customer data points … it’s alllllll going into tools that may store, learn from, or surface that information in ways that no one fully controls.

Meanwhile, communicators are being asked to explain AI externally to investors, customers, employees, and the media right at the moment regulators are paying very close attention.

The absence of clear policy doesn’t create caution. It creates chaos. One person treats AI like a spellcheck. Another treats it like a ghostwriter. A third pastes in the board deck without thinking twice. And then the communications team is left trying to close the gap between what the company thought was happening and what was actually happening.

This is bigger than one policy memo

Before we go further, it’s worth naming something plainly: “AI policy” is not one question. It’s a stack of them.

👉 There’s internal use: Can employees upload proprietary documents, customer information, or executive talking points into public AI tools?

👉 There’s IP and confidentiality: How is that information stored, learned from, or potentially exposed?

👉There’s external use: How are AI-generated facts verified? How much AI assistance is acceptable in published work? When does disclosure become necessary? And who’s accountable when something goes sideways?

This isn’t just a technology issue. It’s communications, legal, data, brand, and governance—all tangled up together.

Why corporate comms belongs at the table

From what I’ve read, AI policy typically gets made this way: Legal drafts it. IT reviews it. Maybe HR adds a training module. A PDF gets circulated. A Slack message goes out. Leadership exhales and assumes the company is covered. Done!

But policies don’t change behavior, understanding policies does. And understanding doesn’t come from legal language, it comes from communication. An AI policy lives or dies based on whether employees actually understand what it means for their work, on a Tuesday, with a deadline.

The strongest companies are building cross-functional AI governance teams that include legal, IT, security, HR, risk, and corporate comms. Because someone needs to translate risk into real-world scenarios, policy into day-to-day decisions, company values into actual AI behavior, and confusion into clarity people can act on.

You can mandate tools and restrict platforms and publish gorgeous policy documents all day long, but you cannot enforce what people don’t understand or don’t believe in. And right now, in a lot of companies, there’s the official AI policy and then there’s how work actually gets done. Comms is the function that closes that gap.

A policy starter kit for communicators

So what should a usable corporate AI policy actually do? At a minimum, it should reduce ambiguity in three places: what employees can input, what they can publish, and what the company is willing to claim.

✅ With that in mind, here’s a working checklist I put together to consider:

1. Define what “AI” actually means inside your organization. Don’t let the term stay vague. For context, the SEC’s Investor Advisory Committee recommended that companies define what they mean by “artificial intelligence” in disclosures and distinguish between low-risk support tools (like transcription or grammar help) and AI that materially shapes products, messaging, or decisions. Your internal definition should be just as specific.

Now, if you’re not at a publicly traded company, you might be thinking SEC enforcement actions don’t apply to me. Fair. But the principle does. Regulators are establishing that vague or inflated AI claims carry consequences. That bar is going to trickle into every sector — healthcare, education, nonprofits, government, retail. If your organization is saying anything publicly about how it uses AI, accuracy and consistency matter.

2. Use one AI narrative across audiences. If leadership is talking about AI on earnings calls, the company should be equally clear in PR, marketing, recruiting language, and executive remarks. “AI” can’t be a buzzword in one venue and a carefully qualified capability in another. The SEC is flagging inconsistency and vagueness not just misstatement.

3. Decide when disclosure is required. The IAB (the Interactive Advertising Bureau sets the standards for digital advertising) put out an AI Transparency and Disclosure Framework that I think is worth knowing about, even if you’re not in advertising. What I like about it is that it doesn’t say “slap a label on everything that touches AI.” Instead, it asks a more practical question: Would a reasonable person be misled by not knowing AI was involved here? Would it change how they understand what they’re looking at? If yes, disclose. If not, move on. That’s a framework communicators in any industry can borrow.

4. Write rules for drafting, originality, and verification. Spell out where AI can help and where humans must lead. Brainstorming, summarization, and first-draft assistance may be fine. Final factual verification, executive voice, strategic judgment, and regulated communications should stay clearly human-accountable. Generative AI can produce false citations and flat-out invented facts (the industry calls them “hallucinations”)which is a generous word for “confidently making things up.” Courts and professional bodies have already sanctioned people for relying on unverified AI output, so that’s not just a theoretical risk.

5. Restrict sensitive inputs. A real AI policy should answer the question employees face every single day: What can I paste into this tool, and what can’t I? Confidential documents, whistleblower information, unpublished financials, customer data, and proprietary strategy need very clear rules.

6. Establish review and escalation paths. If AI-assisted output is headed for investors, customers, regulators, the media, or the public, it should trigger clear human review. Define who signs off, who verifies claims, and who gets called when something goes wrong.

7. Make accountability visible. The SEC’s advisory committee explicitly recommended disclosure of board oversight mechanisms for AI deployment. Even if a company doesn’t emphasize this externally, internally it should be clear who owns AI governance, who monitors risk, and who updates policy as things evolve.

8. Build disclosure into the brand system. For companies using AI externally, disclosure language should live in style guides, byline rules, image credits, chatbot labels, and customer-facing user experience. It should not be invented campaign by campaign and under deadline pressure.

Subscribe now

Who’s actually doing this

So, let’s look at a few organizations approaching AI policy as more than a compliance exercise because the range here is worth paying attention to.

🗞️ The New York Times: Human-led journalism with AI guardrails

The Times’ newsroom guidance makes the editorial line clear: generative AI can assist journalists, but the work must remain managed by and accountable to humans. It should begin with vetted factual information and be reviewed by editors. There are also guardrails around sensitive materials, including confidential sources.

Theirs’ is a useful legacy-media model: AI supports the process, but it doesn’t replace editorial accountability.

⭐ For communicators, the takeaway is that even in organizations where speed and volume matter enormously, the policy centers on human judgment as the non-negotiable.

🖥️ Microsoft: Responsible AI as an operating system

Microsoft has built one of the most public-facing responsible AI frameworks out there, grounded in fairness, reliability, safety, privacy, inclusiveness, transparency, and accountability. What’s notable is that it’s not a side memo, it’s embedded into how products are built and deployed. Their 2025 transparency report makes governance visible, not just aspirational.

⭐ For communicators, Microsoft is an example of AI policy as infrastructure, not afterthought.

🎯 Target: The gap between the AI story and the AI policy

Target, however, is an interesting example because it shows the gap in AI talk and AI action.

While the company has made AI central to its growth story with its Store Companion for associates, internal GenAI tools for HQ teams, Trend Brain for merchandising, and its press releases and blog posts emphasizing a ‘safe and responsible use’ and empowering team members rather than replacing them. It doesn’t, however, seem to have a detailed public AI policy that spells out how employees may and may not use these tools day to day.

⭐ That contrast — heavy public emphasis on AI as a capability, lighter public detail on AI as a governed practice — is exactly the kind of gap communicators should be watching for in their own organizations.

The Bottom Line

AI policy shouldn’t be a static document. It’s a living system of decisions, behaviors, and tradeoffs happening across an organization every day.

AI use is no longer invisible. It is increasingly subject to scrutiny from regulators, the media, and the public. And when that scrutiny comes, companies will need to do more than point to a policy. They will need to explain their decisions, defend their practices, and demonstrate consistency. Corporate communications sits at the center of that work not just translating policy, but in helping ensure it holds up in the real world.

Your 5-Minute Challenge

• Open your company’s AI policy (if one exists). Can you explain it to a coworker in a few sentences? If not, that’s your gap.

• Ask one colleague this week: ‘Do you know what you’re allowed to paste into ChatGPT at work?’ The answer will tell you everything.

** Got more time? If your company doesn’t have an AI policy yet, or maybe has one that nobody reads or understands, take a first pass at drafting one. 💪

Use the starter kit above as your outline, keep it to one page, and then invite the people who should be in the room: legal, IT, HR, leadership. You don’t have to have all the answers, but you can be the person who gets the right people talking.

Does your company have an AI policy, and, better yet, do you understand it? Drop it in the comments below, and let’s discuss.

Leave a comment

This week, I gave Microsoft Copilot a real workout inside the tools I use every day. My work runs almost entirely on Microsoft 365 — Outlook, Teams, SharePoint, Excel, PowerPoint — so the test lab was already built.

I used it across my actual day-to-day: drafting executive talking points, refining blog copy and headlines, summarizing and writing emails, building a PowerPoint deck, and reviewing budgets.

What followed was … shall we say, educational. 🥴

Frequent timeouts. Multiple error messages. Somewhat promising outputs that then cut off mid-thought, from a tool that’s supposed to be seamlessly woven into my workday.

Which, fine. New tools are glitchy, and I am someone who documents AI failures in a newsletter for sport.

But here’s what stuck with me—A lot of companies are betting on Copilot as their only AI solution.

And that’s the part worth talking about.

First — what’s Copilot?

Before we go further, it helps to know that ‘Copilot’ is really a brand umbrella but there are several versions, and they’re meaningfully different.

• Microsoft Copilot (Free) — The basic version that’s web-grounded AI chat for general questions and research. You’ve probably seen it inside Bing or Windows. It’s good for casual use but doesn’t offer integration with your actual work files.

• Microsoft Copilot Pro — The paid personal upgrade. It provides faster responses, priority access, image generation, and Copilot integration inside your personal Microsoft 365 apps.

• Microsoft 365 Copilot (Business) — This is the enterprise one, and the version most companies are deploying. It plugs directly into the Microsoft 365 suite and draws on your company’s internal data, so it can actually reference your files, emails, and meetings to give you relevant answers.

Now, let’s talk about the pricing options for Copilot because it’s quite complicated.

What Copilot actually costs (it’s complicated)

The Copilot license is an add-on. It does not replace or include your Microsoft 365 subscription, it sits on top of it (so you’re paying for both).

Here’s how the tiers break down:

• Copilot (Free)/ Best for inviduals, offers basic web chat

• Copilot Pro ($20 per user/mo)/ Best for individuals on persal Microsoft 365 plans

• Copilot Business($21 per user/mo)/ Best for small businesses (under 300 users)

• M365 Copilot Enterprise ($30 per user/mo)/ Best for larger businesses (over 300 users)

• Copilot for Sales ($50 per user/per mo or $20 add-on if you have M365)/Best for sales and customer service teams

To put this in real terms: a mid-sized company with 100 employees paying for both Microsoft 365 Business Standard (~$12.50/user/month) and the Copilot Business add-on ($21/user/month) is looking at $33.50/user/month or about $40,000/year before you’ve even looked at whether people are actually using it. That’s not chump change.

According to a Recon Analytics survey of more than 150,000 U.S. workers, the voluntary usage rate among employees who have access to Copilot sits at about 36%. For ChatGPT, that number is 83% (for context).

Forty thousand dollars a year for a tool that roughly two-thirds of your employees choose not to open when given the choice is definitely something to think about. 💸

Subscribe now

The case for it makes sense … sort of

I want to be fair here, because what Copilot promises to do, app by app, is genuinely impressive on paper:

• In Word: Draft, edit, summarize, rewrite

• In Excel: Identify trends, generate insights, create visualizations

• In PowerPoint: Generate slides, pull content across applications

• In Teams: Summarize meetings, capture action items and decisions, generate transcripts, highlight key moments so you can catch up on what you missed

• In Outlook: Draft emails, summarize long threads, suggest replies

It also learns your prompts, tracks your interactions, and is supposed to adapt to your brand voice and preferred ways of working over time.

And the case for companies going all-in is legitimately clean: tools your team already uses (✔️), data that stays inside your ecosystem (✔️), one vendor (✔️), and one contract (✔️). Too, for companies just trying to get their arms around AI without letting employees run wild with twelve different subscriptions, this makes so much sense.

But, here’s what I’m actually seeing both in my own testing and in researching this.

Copilot works well as a productivity assistant for the routine, repeatable tasks like summarizing a document, tidying up a draft, or pulling quick insights from a file you already have open. But when you ask it to do work that actually moves the needle — think strategy planning, narrative building, nuanced positioning — it starts to fall apart both conceptually and technically.

I’m not alone in noticing this. Multiple enterprise analyses have documented what they call five recurring Copilot limitations:

• Accuracy issues compared with other leading models,

• Inconsistent integration with its own Microsoft 365 apps

• Reliability problems

• Weak admin controls

• Lack of long-term memory

Enterprise IT guides now recommend building workflows that explicitly work around Copilot’s weaknesses rather than assuming it can handle everything, which feels like a lot for a tool that you’re paying $30 per person per month for, right? 🤨

Even Microsoft’s own CEO Satya Nadella reportedly described some Copilot integrations as “almost unusable” internally. Copilot’s accuracy Net Promoter Score (a measure of whether users trust its outputs) hit -24 (!) at one point in 2025. This means that more people were actively warning others away from using it rather than recommending it.

What companies are really hoping for

But looking under the hood a bit, let’s be honest about what may be driving the “Copilot only” mandate. I’m guessing many companies aren’t just adopting AI, they’re trying to contain it. When you factor in the cost of scaling multiple tools and the very real liability concerns around data privacy and what employees are actually putting into these systems, I get it. Considering all of this, a controlled single-vendor solution starts to sound pretty appealing. They want it to be the solve-all, single approved tool (security and compliance sorted); the cost-controlled solution (no subscription stacking); and, the default workflow layer (no behavior change required from employees). But, the one AI to rule them all ideal, while completely understandable, is not how it’s playing out. In fact, there’s a split developing inside many organizations that looks like this:

Copilot is the official, sanctioned tool. But, behind the scenes, people use ChatGPT or Claude to fill in the gaps, not to be rebellious, just to keep the work moving.

There’s even a name for it: Shadow AI, when employees use better-suited AI tools under the radar when the approved ones don’t cut it. And, it’s happening everywhere, at every company size.

The model that’s actually working

The companies getting this right aren’t forcing a single tool. They’re using a layered AI stack like this:

• Copilot = embedded workflow support (for the stuff it genuinely handles well e.g. meeting notes, email drafts, document summaries, data visualization in Excel)

• ChatGPT / Claude = strategy, ideation, complex writing, anything requiring real nuance

Different tools for different kinds of work.

Because “AI” is not one thing, and pretending otherwise creates the exact problem companies are trying to avoid by having employees go around the system anyway, just less visibly.

Why this matters for those of us in Comms

This isn’t just a tooling conversation. It’s a misunderstanding of what AI is actually for.

If organizations treat AI as a cost-saving measure, a procurement decision, or an IT rollout, they’ll miss the bigger shift that AI is changing how thinking work gets done. And that requires flexibility, experimentation, and trusting people to find the tools that help them do their jobs well. (Note: I still very much advocate for human-led critical thinking, deep and soulful creative work, and real-world storytelling).

I’ll speak for myself here: the tasks where Copilot helped me most were the ones that didn’t require much judgment such as cleaning up a draft or pulling a summary from a long email. The moment I needed real strategy or positioning, I was on my own.

The Bottom Line

Copilot is a powerful idea. The app integrations are real, the use cases are legitimate, and Microsoft is investing enormous resources into making this better. The new SMB pricing tier shows they’re hearing the feedback.

But right now? At $21–$30 per user per month on top of your existing Microsoft 365 bill, with a 36% voluntary adoption rate and an accuracy score in negative territory, it’s not enough to be your only AI solution. The companies that acknowledge that early—and give teams the flexibility to build a real AI toolkit—are the ones that will move faster.

Are you navigating a “just use the company-approved AI” situation at work? I’d genuinely love to hear how you’re handling it or creatively not handling it. 😉

Share it here!

Leave a comment

Welcome back!

First, I just want to say thank you to all of you who have subscribed. I’m truly humbled you’re here, and I’ll do my best to keep you interested, informed, and maybe a bit entertained. Heck, that’s what I’m hoping to do for myself these days.

This week I’m looking at AI research tools (and there are many, whew). But, I’ve pulled together the ones that actually move the needle for communications professionals. Also in today’s post, plot twist, I now have a Beta Lab failure to report 🫣, but like any good “scientist” (ok, this metaphor is feeling cringey, apologies), I kept at it and it ends with a win! And I close out with a quick update on the Anthropic vs. Department of Defense saga that I touched on last week because it’s heating up.

🕵🏻 Part One: The AI Research Tool Breakdown

If you work in communications (PR, marketing, public affairs, content strategy), you’ve probably asked yourself: Do I really need anything beyond ChatGPT, Claude, or Gemini for research?

The answer is messy.

It’s mostly no, but sometimes (emphatically!) yes. And figuring out that boundary line is basically the entire struggle. Knowing which tool is the right tool is the whole game right now. So, here’s an honest look at five tools I’ve been testing, and how I actually map them to my workflow.

1. Perplexity (The Librarian)

What it is: What it is: An AI search engine for a ‘post-trust’ world. It pulls from the live web and provides immediate, clickable ‘inline’ citations so you don’t just have to trust that the AI isn’t hallucinating (see below).

Think of it as Google plus a smart summarizer that actually cites its work. It searches in real time, which means it can tell you what reporters are writing about right now, not what was trending in its training data from six months ago.

Side bar: hallucinating is AI jargon for when a tool confidently makes up a ‘fact’ that sounds true but is actually total fiction.

Why does it do this? Because AI doesn’t actually ‘know’ facts; it’s just a world-class pattern matcher. Think of it like a super-powered version of the predictive text on your phone. It’s not checking a library; it’s just guessing the next most likely word in a sequence. If it doesn’t have the truth, it simply predicts what a ‘truth-sounding’ sentence would look like and serves it up with total confidence. And, because AI is programmed to be a ‘document completer’ designed to never leave a sentence unfinished, when it hits a gap in its knowledge, it doesn’t have a ‘fact-check’ button. It only has a ‘what-word-comes-next’ button. So, it chooses the most probable word based on the billions of sentences it has read, which often results in a perfectly grammatical, highly confident, and completely made-up answer. 🤡

Best for comms pros:

• Fast competitive and trend research

• Pulling fresh stats for decks, posts, and pitches

• Finding recent media coverage on a topic

The catch: It can give shallow summaries on complex topics, and you still need to sanity-check the stats it pulls. It’s a research starting point, not a research endpoint.

Pricing: It has a strong free tier; advanced features require Pro which is $20 per month/$200 annually. (There are also enterprise level pricing options)

2. NotebookLM (The Executive Assistant)

What it is: Google’s AI tool (powered by Gemini) that lets you upload your own documents and then chat with them.

This one is less about finding new information and more about making sense of information you already have. Upload a policy brief, a 40-page industry report, a pile of interview transcripts and ask it to synthesize key themes, pull quotes, or build a messaging framework.

Best for comms pros:

• Turning research piles into talking points

• Summarizing client materials or stakeholder interviews

• Building internal knowledge hubs

The catch: It only works with what you give it. It’s not a discovery tool. And, it’s tied to the Google ecosystem, which may or may not be your thing.

Pricing: If you already have a Google account, you have access to this for free. Pro level is $19.99/month but you can’t buy it alone. It’s bundled into the Google AI Pro plan. (There are also enterprise level pricing options)

3. Elicit (The Lab Assistant)

What it is: An AI research assistant built specifically for academic literature reviews, pulling from Semantic Scholar’s database of papers.

To understand Elicit, you first have to know about its engine: Semantic Scholar.

Think of it as a massive, gated library of over 200 million academic and scientific papers. Unlike a standard Google search that pulls from the ‘live’ (and often messy) web, Semantic Scholar only indexes peer-reviewed research.

Because Elicit is built on top of this database, it’s not just ‘guessing’ an answer based on the whole internet; it’s specifically looking for answers within published, verified studies. If you’re looking for the actual science behind consumer behavior or the psychological impact of AI, this is where you go to find the receipts.

If you’re building an evidence-heavy piece, such as a whitepaper, a hero deck, a “state of the industry” report, and you need real studies to back up your claims, Elicit is your friend. It finds relevant papers and extracts key findings in a structured way.

Best for comms pros:

• Building “evidence packets” to support performance or sustainability claims

• Backing up thought leadership with peer-reviewed data

• High-stakes content where “I read it somewhere” won’t cut it

The catch: This is not a daily-use tool. It’s slower, more academic in feel and best saved for projects that actually need that level of rigor.

Pricing: It’s a Workflow-based structure. This means that while basic actions are unlimited, the heavy-duty automated research reports are capped by tier. If you’re just looking to see if there’s any science behind a new consumer trend, the free tier is actually plenty. But if you’re building a white paper or a high-stakes pitch and need to export a table of ‘proof’ to show a client, the $10 Plus tier is your best bet. (There are also enterprise level pricing options)

4. Consensus (The Scientific Jury)

What it is: A search engine specifically for scientific research. It answers questions by aggregating findings across peer-reviewed studies.

Good for a quick gut-check: Does the research broadly support this claim, or is it more complicated than that? It’s like asking “what does science say about X?” and getting a summary instead of 47 conflicting abstracts.

If you’re asking what is the difference between Elicit and Consensus, here’s the scoop: Consensus tells you “What is true.” Elicit tells you “How it was proven.”

Think of Consensus as your high-speed fact-checker for when you’re in a meeting and need to know if a claim holds water. Elicit is who you call when the meeting is over and you actually have to write the ten-page report with the citations to back it up.

Best for comms pros:

• Validating the angle behind a campaign claim

• Quickly checking whether there’s broad consensus or active debate on a topic

The catch: Some of the more powerful features are behind a paywall. And like any tool that aggregates, results still need a manual check.

Pricing: The free tier is great for casual fact-checking and gets you unlimited basic searches across their 200M+ paper database. The Pro Tier is $120/year or $15/month and unlocks unlimited pro searches, snapshots, “ask paper” messages wher eyou chat with a specific PDF, and 15 deep searches per month. (There are also enterprise level pricing options)

5. Scite (The Private Investigator)

What it is: A citation intelligence tool that shows not just how many times a paper was cited, but whether later research supported, contradicted, or simply mentioned it.

This is the “stress test” tool. Before you build a campaign around a single compelling study, Scite can tell you whether that study has held up or whether a dozen papers came out later saying “actually, no.”

Best for comms pros:

• Verifying the credibility of a key study before anchoring messaging to it

• Avoiding the embarrassment of citing a paper that’s since been disputed

The catch: It’s a very narrow use case. It’s not conversational and not great for exploratory research. It’s very academic-heavy. Unless you are involved in high-stakes medical or legal comms, knowing if a paper was “contrasted” by another peer-reviewed study might be more detail than you actually need.

Pricing: Scite is a premium-first tool. Unlike the others, they don’t have a robust forever-free version, but they do offer a 7-day trial. Individual pricing is $20/month or $200/year and this unlocks the Scite Assistant, Full-Text Search, and the ability to see the “Supporting/Contrasting” breakdown for any paper. (Enterprise pricing is customized.)

🤔 So What Should You Actually Use?

For most communications work, the stack that covers you is simpler than you’d think:

• Perplexity for everyday external research and idea generation.

• NotebookLM for making sense of materials you already have.

• Elicit + Consensus + Scite for when you’re building something that needs to stand up to scrutiny.

The general-purpose tools (Claude, ChatGPT, Gemini) are still your workhorses for drafting, reasoning, and synthesis. The research tools offer an additional layer you can add in when you need verifiable, sourced information, not just a well-written summary.

Subscribe now

🧪 Part Two: Beta Lab — The One Where I Got It Wrong (And Then Got It Right)

The Fail

I’m human (a quite determined one, though 😉). And last issue’s Beta Lab did not go as planned.

The experiment was to automate a work industry news briefing using Zapier + ChatGPT + Gmail. The zap ran. The email arrived. But instead of an actual briefing full of news, I got the prompt template I’d built. The skeleton. The instructions for what a briefing should contain, with zero actual content inside it.

Good times.

After a lot of digging, I found out why: Custom GPTs inside ChatGPT cannot be directly triggered by Zapier. They live inside the ChatGPT interface only. So when Zapier runs a “Conversation with ChatGPT” action, it doesn’t open your custom GPT, it ignores it entirely, even if you’ve written identical instructions. So I was essentially asking Zapier to run a blueprint and call it a house.

SO … Back to the lab. 🏃🏻‍♀️

⭐ The Fix

Goal: Automate a daily industry news briefing delivered to my inbox by 8am each weekday.

Tools: Perplexity (paid) + Zapier (paid) + Gmail

The Experiment: This time I started completely fresh with a new approach, new eyes, new AI to walk me through it. I used Gemini to guide me through the build step by step, which felt slightly absurd (using one AI to fix a different AI’s mess), but here we are.

The key insight: instead of feeding news sources into ChatGPT, I switched to Perplexity as the AI engine inside Zapier. The difference is significant. With ChatGPT, you have to manually(!) list every outlet and weblink that you want it to scan. With Perplexity, you just tell it what type of industry news and terms you’re interested in it pulling, and it goes and it does all of the searching itself.

The three-step zap:

1. Schedule by Zapier — triggers the automation each weekday morning

2. Perplexity via Chat Completion — goes out, finds the news, formats the briefing

3. Email by Zapier — delivers it to my inbox

   

Real talk: this setup took hours. There were many test failures, many screenshots uploaded to Gemini for troubleshooting, and many more than a few moments of staring at my screen wondering what I was doing with my life. But I got there. 💪🏻

Verdict: Worth it — but budget serious time for the learning curve if Zapier isn’t your thing. Once it’s running, it’s genuinely useful. And yes, both Perplexity and Zapier paid plans are required for this setup. Zapier does offer a 14-day free trial, so you can test before you commit.

(Screenshot of the working zap included above — yes, it’s real, yes it works!)

⚠️ Part Three: Anthropic vs. The Pentagon

Remember the drama I mentioned last week regarding Anthropic and the Department of Defense? Well, things just went from “tense” to “litigious.”

On Monday, Anthropic filed two major lawsuits challenging the government’s decision to label them a “supply chain risk.” To give you some context, that’s a label usually reserved for adversarial foreign companies, not a home-grown AI startup in San Francisco.

The move came after President Trump directed federal agencies to stop using Anthropic’s AI, effectively banning them from defense systems. Anthropic’s response was direct:

“This is a necessary step to protect our business, our customers, and our partners. We will continue to pursue every path toward resolution, including dialogue with the government.”

The “Rivals” Step In

What I found absolutely fascinating—and frankly, a bit hopeful—is that on Monday, 37 employees from OpenAI and Google (including Jeff Dean, Google’s Chief Scientist) filed a legal brief supporting Anthropic.

Seeing competitors set aside the “AI arms race” to stand together against the Defense Department is a huge signal. They aren’t just defending a rival; they’re defending the right to have guardrails. They cited their own understanding of the risks of “frontier AI” and the need to prevent things like domestic mass surveillance or fully autonomous weapons.

The Integrity Question

As I’ve said before, I admire Anthropic’s willingness to stand up to the US government to protect their safety ideals. It’s a massive “David vs. Goliath” moment.

However, in the spirit of keeping it real, their internal culture is also under fire. Their head of Safeguards Research, Mrinank Sharma, recently resigned with a deeply unsettling open letter. He warned that “the world is in peril” (Dear God) and admitted how hard it truly is to let values govern actions within an organization facing this kind of pressure. 😱

I don’t want to be a “Debbie Downer” or get too far into the weeds of this always-terrifying news cycle, but staying informed is the only way we get to make our own decisions. (Don’t even get me started on the environmental impact of these data centers—that’s a post for another day for sure. 😔)

Leave a comment

Your 5-Minute Challenge:

Pick one of the tools above (most have free tiers) and try to find one ‘fact’ you’ve been using in your pitches that might actually be a hallucination.

If you find one, reply to this email and let me know. I’d love to feature it in an upcoming post.

Someone at a recent communications conference I attended said something that made me sink a little in my chair with the feeling oh no, I should already be doing this.

“If you do a task more than twice, you should be automating it with AI.”

So this issue is dedicated to that concept because if you’re not building custom AI tools for your repeat tasks yet, you’re leaving time on the table. And time, as my two kids remind me every single day, is not something I have in surplus.

🔧 Custom GPTs: Your AI, Trained to Think Like You (Sort Of)

First, what is a GPT? I personally love how everyone throws around AI terms like we all got the same memo. 🤦🏻‍♀️ We didn’t, and I’m willing to admit it.

GPT stands for “Generative Pre-trained Transformer.” That’s a mouthful, so let me break it down: It’s essentially an AI that read the entire internet, learned from it, and now writes back to you based on what you ask it. ChatGPT is the one everyone knows, but think of GPT as the family name with ChatGPT being the celebrity in the family.

A Custom GPT is essentially a version of one of these models, like ChatGPT, that you set up with your own specific instructions, personality, or knowledge set. Instead of starting from scratch every time you open a chat or task window, your custom GPT already has context and pre-loaded knowledge pertinent to what you’re looking to do. It basically already knows the drill.

Just think of the tasks you consistently do on repeat:

• Writing social posts for the same campaign types

• Drafting executive email updates that always follow the same structure

• Summarizing research into the same format every week

• Generating content ideas for your editorial calendar

A custom GPT can handle the first draft of each of those tasks. Of course, you still want to be the human in the mix who shapes it, edits it, and gives it fresh non-robot life (remember, we’re anti AI slop), but by using this, you’re not starting from a blank page every time.

Where can you build one?

• OpenAI’s ChatGPT: the most intuitive option. You can build a custom GPT through their “Create GPT” feature.

  I tried this for an industry headlines digest and it was shockingly easy.

  ⭐ Pro tip: First, you must be using a paid subscription in order to build a custom GPT. Then, if you don’t see "Create” in your “Explore GPTs” section, just type in this url — https://chatgpt.com/gpts/editor — into your browser and it will take you straight to it.

• Claude Projects: you can build a project with custom instructions and uploaded documents that shape how Claude responds to you within that context. I use a Claude Project for creating this substack.

• Google’s Gemini Gems: similar concept, Google’s flavor.

• Microsoft CoPilot Agent Builder: and here’s where I need to stop and be honest with you.

I tried to build a daily news briefing through CoPilot, but it first told me it couldn’t pull from the web. Oddly, when I refreshed it, it did pull information but the articles it summarized didn’t actually exist because it didn’t include specific weblinks. How I discovered this was that I typed in the headlines and news source that was sited in the briefing, and when I searched for them, they weren’t there. It was all fabricated (a “hallucination” in AI lingo).

Too, it kept interrupting me by repeatedly throwing up a window asking me to save this command into an SVG file (still have no idea). Finally, when I came back to refine it the next day thinking maybe a breather would do us all some good, it had zero memory of anything I’d done.😡

If your company runs on Microsoft and CoPilot is your mandated AI tool, I feel for you. I really do.

A hack worth stealing:

Use ChatGPT to help you design your custom GPT. Prompt it using something similar to this (or your version of what you’re trying to achieve):

“Help me define the tone and content guidelines for a custom GPT that writes newsletter content for a B2B brand in the [X] industry. The tone should feel [X], length should be [X], and the audience is [X].”

Then use those guidelines when you build your GPT. Let the AI help you train the AI.

You can also upload documents such as past emails, reports, interview transcripts, and brand voice guides to give your GPT real context to work from. The more you feed it, the better it gets at nailing down what you need.

🧪 Beta Lab: I Automated My Morning News Briefing (And Only Wanted to Quit Once)

Goal: To stop manually hunting for industry news every morning and replace it with an automated daily briefing delivered to my inbox each weekday morning by 8am.

Tools: ChatGPT (custom GPT) + Zapier (free plan) + OpenAI API

The Experiment: I built a custom GPT for daily industry news summaries and then used Zapier to automate it so that every weekday at 8am, an email titled “Daily News Briefing” lands in my inbox without me lifting a finger. The idea was straightforward. The execution was ... a journey.🤯

I used ChatGPT to help me build the Zap from scratch, including the exact language to describe my workflow: “Every weekday at 9 AM, generate a daily news briefing about U.S. industrial hemp fiber, hemp textiles, sustainable materials, and manufacturing reshoring policy, then email the briefing to me.”

It told me which trigger app to select (“Schedule by Zapier”), which action app (“OpenAI”), and which action event (“Send Outbound Email”). Essentially it handed me a map before I even opened the door.

The tricky part? Connecting the steps required my OpenAI API key — which I didn’t have, because I didn’t know it existed. Turns out, your ChatGPT subscription and your OpenAI account are completely separate things. Circling back up to what I mentioned before (and what is slowly seeping into my brain cells), ChatGPT is the chatbot and OpenAI is the underlying infrastructure. Building automations like this requires billing through OpenAI directly, on a pay-as-you-go basis.

I also hit a wall when Zapier’s AI assistant, Zapier CoPilot, told me I needed to upgrade to a paid plan to build this Zap. I didn’t want to upgrade, so I took a screenshot and dropped it into ChatGPT. Its response: “What Zapier’s Copilot told you is technically correct but misleading for your use case. The trick is: combine the generation and the email in one step so the Zap only has 2 steps (which works on the free plan).”

So, I restructured the Zap without upgrading per ChatGPT’s instructions, and it worked (high five, ChatGPT!)

As for the OpenAI billing, I loaded $10, which buys “tokens” (yet another AI term we’re all just supposed to know. They are small chunks of text the AI processes including words, punctuation, parts of longer words). How far will the $10 get me? Oh boy, so the AI token discussion is genuinely complicated and deserves its own post. But ChatGPT assured me it’ll power a lot of morning briefings, and I’m choosing to believe it.

Verdict: This is absolutely worth building, but budget time for the learning curve if you’re not familiar with Zapier set up. The free tools are genuinely free if you know how to structure things correctly. When you get stuck, use ChatGPT to troubleshoot.

Subscribe now

🎭 Personas: The Audience You Build Before You Write a Word

This one also came out of the conference and it clicked for me immediately because it’s actually not a new concept. It’s a classic communications practice dressed up in AI clothes.

What’s a persona? In marketing, you’ve probably built customer personas, or detailed profiles of your target audience, before. Think “busy working mom,” “budget-conscious small business owner,” “retired army veteran.” Whatever archetype represents the person you’re trying to reach, you map out their demographics, their pain points, and their goals, so your message actually speaks to them.

And here’s where I’ll out myself a little because personas aren’t just a marketing exercise. They apply inside your organization too, and I haven’t always been great at it. Your CEO and your frontline manager care about completely different things. Your trade press contact and a local beat reporter are operating in different worlds. We know this, but we often send everyone the same email anyway.

The conference tip: build personas in your AI tools for each of your key stakeholders—both internal and external. In other words, you’ll build a persona for your CEO, your CFO, your immediate boss, or your individual team members. Externally, you can build one for a trade press journalist vs. a general news reporter. Then, when you’re drafting communications, you upload the message into your AI tool and ask it to speak to the target persona.

Here’s a prompt to get you started:

“Create a detailed persona for a [job title] working in [industry/company size] who is responsible for [specific function]. What are their likely goals, communication preferences, and pain points?”

Once you have that persona built, you can use it to pressure-test your drafts. Does this message land for a CFO who cares about risk and ROI? Does this pitch work for a journalist who covers policy, not product?

The payoff is in the output. And here’s where it connects back to everything we just talked about: Build a persona, load it into a custom GPT, and suddenly your AI already knows who it’s talking to before you’ve typed a word. No more writing five versions of the same message. Just one good one, tailored from the start.

⚠️ Anthropic vs. the Department of Defense: The Story I Actually Can’t Stop Thinking About

Okay. Major gear shift. But, it’s important to talk about these major AI news stories.

If you haven’t seen this, Anthropic, the company that makes Claude, published a statement about their newly-fractured relationship with the U.S. Department of Defense this week. And then OpenAI, which immediately jumped in to fill the void that Anthropic left when their contracts with the Department of Defense dissolved, had to walk back an agreement they’d made due to customer backlash. (More on this below)

Here’s why this matters to me, beyond just the headline.

We’re in this moment where AI is being built and deployed at a speed that is genuinely hard to track. And most of the conversation in professional circles is still about: can it write my email? Can it summarize my meeting notes? Which — yes! That’s literally what this newsletter is about.

But the bigger conversation happening in parallel is: who decides how this technology gets used? And for what? (Check out this earlier post, Who’s Afraid of Little ‘Ole AI?)

I have a lot of respect for the fact that Anthropic built their entire company around AI safety as a foundational principle, not just a footnote. When a company that makes AI tools says publicly, here are the lines we won’t cross, and then actually holds that line when a government agency comes knocking, that’s not nothing.

Mass surveillance. Fully autonomous weapons. These aren’t science fiction scenarios anymore. These are real conversations happening at the highest levels of government and industry right now, while most of us are just figuring out how to use AI to draft our Monday recap emails.

I’m not here to tell you what to think about the government’s role in AI. But I am here to say: pay attention. The companies building your tools have values, both stated and unstated. And, it’s worth knowing what they are.

If you want to go deeper, read the actual Anthropic statement here and the BBC’s coverage of the OpenAI reversal here. Read them, form your own opinion, and share it below. I genuinely want to know where you land with this?

Leave a comment

Your 20-Minute homework

Three takeaways you can actually use this week:

• Pick one task you do on repeat and spend 20 minutes building a custom GPT or Claude Project around it. Start stupidly simple.

• Build one persona for a key stakeholder using the prompt above. Use it to audit the next communication you send them.

  Thanks for reading You Beta Believe It ! This post is public so feel free to share it.

  Share

My kids don’t believe anything anymore. Not a video, not a photo, not a heartwarming clip of a duck befriending a beagle. I get their signature side-eye and a quick “that’s AI, Mom,” every time. Meanwhile I’m over here living my best life sharing blooper reels and baby videos like it’s 2014. I want to say they’re too cynical. But honestly? I’m not sure I can tell the difference either. 🤔

And that scares me a little.

Sure, I’m getting better at spotting the signs (in written content; clearly, I struggle mightily with discerning AI vs. authentic visual content): the suspiciously tidy structure (think five-point paragraph style which I happen to love, just as I love the much-villified em dash) with a few too many formulaic transitions (think “futhermore” and “however”), or the complete absence of a run-on sentence or random, humanistic sideways thought. But really, it’s my gut reaction versus any kind of analysis that I mostly rely on. Over-polished content does the same thing to me as the 7:30am school drop-off moms that arrive in full glam, matching athleisure, and a smile that suggest they’ve already worked out. Can’t relate. I respect it, but I don’t trust it. That’s AI slop in a nutshell.

And here’s the uncomfortable part, a lot of us in communications are producing it (✋). Not because we’re lazy, but because the tools are just so damn good. And after so many years of juggling imminent deadlines with a serious case of writer’s block, well, it’s just too easy.

So What Even Is AI Slop?

As I mentioned, I’m still learning to spot it myself. I use AI tools to help me draft content (including this newsletter) so I’m hardly in a position to throw stones. But I’m starting to recognize the feeling, even when I can’t always name it.

It’s content that’s technically fine but simultaneously completely empty. The press release that says all the right things but makes you feel nothing. The LinkedIn post with five bullet points but zero personality. The pitch that’s clearly been “personalized” by something that has never actually met a person. It’s sanitized, safe, and structured within an inch of its life.

‘Slop’ was coined the 2025 word of the year, and I’ll be honest when I first heard it I felt a little called out. AI tools are supposed to help us communicate better, but what a lot of us are actually doing is communicating more — faster, higher volume, and in a voice that belongs to no one in particular.

The Stuff AI Actually Cannot Do

I use AI constantly. I’m not here to tell you to throw it out. But I’ve been thinking a lot about where it actually falls short, and the answer is basically everything that made you good at this job in the first place.

AI has never bombed a pitch and had to think on its feet. It’s never sat across from a journalist at lunch and laughed hysterically over an unforgettable story of driving home from a grocery store distracted and overwhelmed, with the family’s rotisserie chicken dinner sliding off the roof of the car.

That last one isn’t hypothetical. One of my longest and most valued media relationships — with the wonderfully talented Elisabeth Egan at The New York Times — wasn’t built on a perfectly crafted email or an airtight pitch. It was built over a lunch where she told me exactly that story, and we laughed about it until our eyes watered. That moment of recognizing each other as two actual humans juggling entirely too much did more for our working relationship than any polished outreach ever could have.

No AI is generating that lunch. No AI is building that relationship. And in a world where everyone’s pitch inbox is being flooded by tools that can produce a “personalized” outreach email in about four seconds, that kind of genuine human connection is worth more than it has ever been.

A few ways to protect and grow your human relationship advantage:

• Do the research, but write the opening line yourself. Use AI to build your media list and draft the pitch, then throw out the first sentence and rewrite it in your own words with something specific and real. “I loved your piece on X” only works if you actually read it.

• Follow up like a human, not a sequence. There’s a difference between a thoughtful check-in and an automated drip campaign. Journalists can tell.

• Check in when you don’t need anything. Send the article that made you think of them. Congratulate them on a great piece. Relationships that only activate when you have a pitch aren’t really relationships.

• Think long game. The reporter who takes your call during a crisis is the one you’ve been genuinely showing up for all along. That’s not a strategy, it’s just how trust works.

• Have the lunch. Seriously. Get off email. The rotisserie chicken moments don’t happen in a pitch deck.

Imperfect Is the New Premium — and This Is Going to Be Hard to Accept

I have spent my entire career chasing flawless deliverables. You probably have too. So this next part is going to feel a little wrong before it feels right.

The most credible content in 2026 is going to be slightly messy. A January 2026 Digiday report found that only 26% of consumers now prefer AI-generated content over human-created content — down from 60% just two years ago. Brands are apparently getting the memo too, with some now requesting imperfections from their content creators. Authentic imperfection as a competitive advantage.

This doesn’t mean sloppy. It doesn’t mean unedited chaos. It means let your actual voice show up. Write a post about something you’re still figuring out. Admit when a campaign didn’t land the way you thought it would. Use the weird specific detail instead of the clean general one. Stop editing the you out of everything.

Some practical ways to stay real when you’re working with AI:

• Do a brain dump first- use voice memos, messy notes, a rambling draft, whatever works. Get your own raw thoughts down before you open any tool. Give AI something to work with, not a blank page to fill for you.

• Let AI shape it, not replace it. Use it to tighten, organize, and polish. Use it to help you outline your piece. But, not to actually think for you.

• Feed your AI your brand voice. Before you start any piece, share your brand voice guidelines including tone, terms, tempo, approved terms, terms you don’t approve, and your brand quirks directly with your AI tool. This helps to ensure that the content that the AI tool develops is created through this critical lens.

Storytelling Is the Skill Worth Protecting Right Now

If AI is handling more of the execution, the people who win are going to be the ones with strong enough editorial instincts to direct it. That means knowing what angle is interesting, what detail makes something land, what story is actually worth telling. You can’t outsource that judgment, but you can sharpen it.

A few ways to sharpen your storytelling chops right now:

• Read more long-form journalism. And, not just for work. But, because it’s good!

• Study the pieces that made you feel something. Reverse engineer why they worked. What detail landed? What made you keep reading?

• Take your storytelling seriously as a craft, not just a job function. It is, not to be dramatic about it, the one thing the robots don’t have.

🧪 The Beta Lab: Does an AI Detector Know You’re Using AI?

Goal: Figure out whether AI content detection tools can actually tell the difference between human and AI-generated content, and what that means for communications professionals.

Tool: GPTZero (free version)

The Experiment: In the spirit of this entire newsletter, I fed a draft of this very post into GPTZero — one of the most widely used AI detection tools out there (there are many) — to see what it would say. While this piece was written with AI assistance, it largely consists of my own content, in my own voice.

The Results: It flagged it, specifically for “formulaic organization and transitions.” Clear headings, logical progression, consistent formatting, and “mechanical” sentence connections. Apparently that’s suspicious now. Personally, I like structure, but that now makes me look like a robot.

And that’s where I have real concerns; not for me, but for the student who writes a clean, well-organized essay and gets accused of cheating because their writing is “too coherent.” These tools are working with patterns and probabilities, not facts. Which is fine for a gut-check on your own content. Less fine if someone’s academic future is riding on the result.

Verdict: Use these tools as a signal, not a verdict. The more useful question isn’t will this get flagged, it’s does this actually sound like me? Your stories, your opinions, your specific weird details are your best defense against the detector and your best asset as a communicator.

Your 20-Minute homework

Run one of your own recent pieces through GPTZero (free). If it flags it heavily, don’t panic, just use it as a prompt to go back and inject more of your actual self into the piece. Your stories. Your specific opinions. Your rotisserie chicken moments.

What’s your signal that something is AI-generated? Drop it in the comments.

Image: Midjourney / Prompt by Beth Gebhard

By now, you may have seen it floating across your LinkedIn feed, your group chats, or your anxiety-fueled late-night doom scrolling. Matt Shumer’s piece “Something Big Is Happening in AI — and Most People Will Be Blindsided” — racked up millions of views and set off a collective “wait... what?” moment across pretty much every industry.

I read it. And then I sat with it for a while.

Here’s the honest, unfiltered version of what went through my head.

But first …

Who is Matt Shumer? He’s the co-founder and CEO of OthersideAI, an applied AI company, and someone who has spent six years building in the AI space. He’s not a pundit, he’s an operator. Which is exactly why his piece hit differently. He originally posted it on X, it got adapted and published in Fortune, and it’s now been viewed by millions and millions of people.

The Urgency Is Real. The Panic Is Optional.

Let’s clarify what exactly Shumer is saying, because the headline alone is enough to send someone spiraling:

By the end of this year (or next at the latest) AI is likely to be capable of performing any task a human can do on a computer, at least as well, and in many cases better.

Read that again. Slowly.

Now breathe.

But, here’s the thing, when this hits your industry matters. Diffusion isn’t instant. Some fields are already deep in the AI transformation. Others are still on the edges. But it seems the direction of travel is not up for debate. Shumer recently said in an interview on CBS Mornings that AI is improving at roughly 100x per year— a rate at which no other technology in history has ever moved.🤯

So no, I’m not telling you to panic. But I am telling you that if you’re not actively learning and using AI in your work right now, you are falling behind. Full stop.

My Three Honest Gut Reactions

1. Urgency, but not the sky-is-falling kind. Any professional who isn’t actively building AI skills and working it into their daily workflow is going to get left behind. That’s not a hot take. That’s just the math.

2. Excitement, genuinely. For those of us who still have a LOT of runway ahead. (hi ✋, I feel like I’m just getting warmed up in my career), this moment feels less like a threat and more like a wide-open door. A chance to rewrite the story. To build something new. To become the person in the room who gets this stuff.

3. Fear for my kids. This one hit differently. The hard work of wrestling with an idea until it clicks, of writing something beautiful from scratch, of building something with your own two hands … I worry those experiences are going to be shortchanged for the next generation. Even I notice it in myself. I don’t sit with words and ideas the way I used to. AI helps me harness my thoughts faster, and that’s a welcome help that’s incredibly powerful. But I don’t want my kids to be robbed of the struggle that I’ve had before AI entered the picture. The struggle is where the growth lives.

So. What Are You Going to Do About It?

This is where I refuse to let this post (or this entire Substack, for that matter) become just another think-piece. You read enough of those. Here’s what I’m actually doing, and what I’d encourage you to start doing too.

Step 1: Upgrade to a paid AI tool and actually use it. I moved to Claude Pro ($20/month or $200/year). The free versions are fine for dipping your toes in. But if you’re serious about this, you need the deeper features that include more usage opportunities, longer conversations, and tools like Claude Projects, which let you build focused workspaces with your own documents and context. Think of it as having a really smart assistant who actually remembers what you told them.

Step 2: Think in context, not just commands. I touched in this in my last post (linked here), but here’s something that genuinely surprised me: good communicators are naturally good at AI. Why? Because AI tools need context. The more clearly you can explain a situation, a goal, or a problem, the better your output. So, yay for us! This is literally the skill set we as communicators have been honing for years.

Step 3: Start exploring the ecosystem beyond ChatGPT. Most people only really know ChatGPT. But there’s a whole world out there! For example, there is Notion AI for organization and knowledge management, Perplexity AI for research and real-time answers, Otter AI for meetings and transcription, and tools like BuzzSumo for content intelligence. You don’t need to master all of them overnight. But you need to know they exist and start experimenting.

Step 4: Let AI help you build your own AI game plan. This is maybe the most meta tip I have, but it works (I did it): Ask your AI tool to help you create a personalized learning plan for mastering AI in your specific field. Tell it your job, your goals, your skill level. Let it map the path. Then follow it.

🧪 The Beta Lab: Upgrading to Claude Pro

Goal: Figure out what upgrading from the free version of Claude to Claude Pro ($20/month) gets me.

Tool: Claude (Free vs. Pro)

The Experiment: I’d been using the free version of Claude for a while and honestly thought it was fine. Then I upgraded to Pro this week and started using it for work in drafting content, building out ideas, and asking it to create a personalized AI mastery plan for a midlife comms professional.

The Results: A few things I noticed immediately.

• First, the answers feel like me. Not in a creepy way, but in a “wait, how did it know to say it like that” way. It turns out that Pro unlocks Claude’s more powerful Opus model, which is significantly better at picking up on context, tone, and nuance. On the free version, you’re working with Sonnet which is solid, but not the same.

• Second, the AI mastery plan it generated would have taken me days to research, format, and organize myself. It was structured, broken into phases, tailored to my actual goals, and designed. And, I had it in minutes. That alone justified the $20.

• Third — and this one’s underrated — Pro includes Claude Projects, which lets you build a persistent workspace with your own documents and context loaded in. No more re-explaining who you are every single conversation. It actually remembers. For anyone doing ongoing creative or professional work, that’s a game changer. Now, if you’re using ChatGPT, it has true automatic memory by learning about you across every conversation without you doing anything. Gemini does this, too, especially if you’re in the Google ecosystem. Claude actually starts every conversation with a blank slate. The memory only activates when you explicitly invoke it.

So what does Claude Projects actually do that’s different? It’s less about “remembering you” and more about using what you upload—your documents, your brand voice, your style guide, your past work— and Claude uses all of that as a foundation for that specific project. It’s more curated and controlled than ChatGPT’s automatic memory.

Neither is strictly better, they’re just different philosophies.

Verdict: If you’re dabbling, free is fine. If you’re serious about making AI part of your actual workflow? The upgrade is worth every penny of the $20.

The Honest Truth About Why I’m Writing This Substack

I’m not writing You Beta Believe It because I have it all figured out. I’m writing it because I’m in the thick of figuring it out — in real time, with real stakes, and a healthy amount of sarcasm to keep things fun.

My goal is to become so good at using AI for communications work that I become the person others call when they need help navigating this shift. Not because I went to a bootcamp or got a certification, but because I showed up, rolled up my sleeves, and just did the work.

Anyway. That's why I'm here, and I’m glad you are too.

Your 20-Minute Homework

Build Your own AI Mastery Plan using Claude, ChatGPT, or Gemini. Prompt it to build you a personalized AI learning plan for your specific job.

👉🏻 Here is your exact prompt:

I need you to build me a personalized AI learning plan.

Here’s my situation: I work in [describe your job/industry]. My current skill level with AI is [beginner/some experience/intermediate]. My biggest professional goals right now are [list 1-3 goals]. My biggest day-to-day work challenges are [list 1-3 challenges]. I have about [X hours per week] to dedicate to learning.

Please build me a 90-day AI mastery plan that includes: the specific tools I should learn and in what order, what I should focus on in each month, one or two concrete things I can practice each week, and how I’ll know I’m making progress. Please keep it practical. I’m not trying to become a developer, I just want to use AI to do my job better and faster.

Have thoughts on the Shumer piece? Questions about any of the tools I mentioned? Drop them in the comments, I’d love to discuss!