Frictionless Decisions brings you counterintuitive news and ideas to help you use data for better, faster decisions.

Subscribe now

A daily report (of anything) sounds so old-school these days. You’ve heard so much about self-service analytics, especially with AI, that you might even be embarrassed to suggest publishing any report from your system.

Then again, you might also be surprised at how much alignment a daily, system generated report creates among decision makers. Publishing data to every level of management, delivered to their email inbox at the same time every day, sets a cadence for everyone to follow, like a drum major leading a marching band. It synchronizes not just the pace of data, but also the pace of thought.

Synchronizing thought? I’m not saying that every decision maker will arrive at the same conclusions about the meaning of the numbers. I’m saying something much simpler. A daily report helps everyone who receives it use the same context for comparisons.

Natural Context

Consultants love buzzwords. “Context” might be the term they’ve most abused over the last year. Here’s a simple way to think about it: if someone tells you you’re carrying a lot of weight, they could be commenting on your eating habits or the extent of your responsibilities. Fortunately, knowing context is much simpler in the business intelligence world. It simply refers to the data points you use for comparisons.

The daily report makes that context obvious - most executives compare data to the previous numbers, and now they’ve got lots of consistently defined comparison points. This simple rhythm helps every decision march to the same beat.

Think about the other benefits:

• The daily report eliminates debates about what time of day the data is “as of.” The data hits your inbox at 8 am, just like it did for everyone else.

• It helps the data team recognize data issues sooner and focus on reliability. Delivering data daily gives you 365 chances a year to perfect the process.

• It creates an environment where incremental improvement is normal.

• It enables management to respond to near-term business changes almost instantly.

Here’s a hidden trick about daily reporting: While “real-time” might mean that a metric was updated three seconds ago for some people, to an executive, a daily report is effectively “real-time” because a day is the smallest bucket of time she uses in her thought processes. With daily reports, you get credit for real-time data without sacrificing context.

There’s an obvious catch to this reporting approach: you need to find a report format that all decision-makers accept. If you schedule all kinds of different reports to distribute daily, you could trivialize the data and dilute the overall alignment among decision-makers. My advice is to start by building consensus on a format among mid- to upper-level managers in your company. Senior executives often rely on them for keeping a pulse on the business.

Maybe you’ve noticed the underlying irony about a data cadence: it’s not technical. Data leaders usually go straight to software and system solutions to solve data problems, but getting everyone marching to the same rhythm means aligning people, not just data. Just because you can deliver data in real-time doesn’t mean you should.

As the leader of the band, this is your job. You help people keep pace with good management practices, like daily reporting.

To remind you of this week’s data concept, enjoy Friday, I’m in Love by The Cure, from the Frictionless Data Spotify playlist.

Frictionless Decisions brings you counterintuitive, original, jargon-free news and ideas for connecting data to decisions.

Subscribe now

Why is the most interesting news around AI all coming out of China? I doubt there’s a simple explanation, but the profit-focused US news can get a little boring. Add to that the context of Chinese education culture, and it creates some fascinating stories, like ’s Prove You’re Human below. For my readers working in higher-ed, you’ll find some rich application this month.

These monthly news roundups focus on global technology trends, personal effects of AI, and (of course) data engineering. That’s why I call it “Connecting Data to Decisions”: I want to help you, who work in the trenches managing data, find ways to avoid the negative effects of data technologies in your company and in your soul. We need to ask the obvious questions (can values be reduced to data?) and take them seriously.

I’ve continued to develop some favorite writers in this month’s list. You’ll also find some fun, new voices. These articles below suggest a deeper, more thoughtful approach than most. Digesting all this takes a lot of time (which you probably don’t have), so I’ve started condensing and curating the best articles and posts for you.

Give them a read!

Global Tech Trends

The DeepSeek Thesis by Irene Zhang, ChinaTalk, August 13, 2026.

A leaked transcript of Liang Wenfeng discussing the state of AI generated lots of discussion last month. Reading about it, I have to admit, I like Wenfeng’s style! I’m not sure his goals fit with my values, but the calm, theoretical approach to work is inspiring. I wonder if this pursuit of knowledge over short-term profit ultimately wins the AI race. Maybe it’s not a race at all.

Value Alignment Is a Pseudo Concept, by Wang Huanchao, Until Zilan Finds a Better Name, August 14, 2026.

Can values be reduced to logic? Can truth be quantified? You might be surprised to learn that this is such a hot topic, not in religious circles, but in the boardrooms and magazines of Silicon Valley. “The dominant technical approaches to alignment today…treat the irreducibility of values as a problem to be eliminated.” Great insight. I can’t say that I’d agree with all the points of view in the article, but his five-point argument is hard to deny.

One more fun fact: the author works for Tencent Research Institute, the Chinese tech company building “technologies that create the future.”

This Former Intel CEO Wants to Jumpstart Moore’s Law With Light by Joe Khalili, WIRED, July 21, 2026.

Pat Gelsinger wants to “wake up” Moore’s Law, and who better to do it than him. He thinks reinventing the way transistors are built is the key, not just better semiconductor designs. Photolithography over processor innovation. Here’s the cool thing: it all hinges on how machines use light. Gelsinger says, “Light is the most important thing. God said, ‘Let there be light.’ We’re going to harness that as far as we can take it.”

AI’s Personal Impact

Prove You’re Human by Zilan Qian, China Talk, July 27, 2026.

For decades in China, the skills required for a career as an illustrator were carefully defined and measured by an extensive, state-sponsored education system. Aspiring illustrators worked extremely hard to align their skills with the testing requirements. As a result of this standardization, AI can easily match the quality of most any Chinese illustrator, and this has caused illustrators to search for new ways to prove their work was done by a human. Those proofs really don’t prove much.

Universities would prefer not to by Hollis Robbins, Anecdotal Value, August 6, 2026.

Seems like there’s a rich theoretical space for higher ed. Like, what’s an ontology? Who should control them and how? Why did technologists co-opt this term in the first place? One other thought: perhaps Ed has a greater impact than recognized, especially in light of the NVidia “five layer cake” analogy.

“University leaders should be staging the national argument about what AI means for history, for language, for being a person. Instead, I see universities as roadblocks to progress,” says. Readers, take note!

Treasure Your Attention by Jon Haidt, After Babel, May 20, 2026.

A beautiful piece, reminding us to avoid getting eaten by The Machine.

Data Engineering

Self-serve analytics has been the data world’s longest-running broken promise by Alejandro Lopez, August 6, 2026.

offers a really great list of four unstated assumptions behind Self-Service Analytics. Then he destroys them. He’s right on target. Put the service back in self-service. Also, take a look back at how SeattleDataGuy offered an alternative approach in his recent article: integrate data at the application layer.

Why General AI Models Still Cannot Be Trusted for Data Analysis by Nick Potkalitsky, Educating AI, June 29, 2026.

I wonder if every company is massively overcomplicating the use case for AI in BI. There’s no such thing as hallucination in business intelligence; the answer is either right or wrong. Data analytics differs from other AI use cases because it requires precision, freshness, and standard business definitions.

summarizes every AI analytics test: “We are not in a crisis of AI capability. We are in a crisis of deployment, moving far faster than our ability to govern what these tools are doing with data that informs real decisions.” Exactly.

From monolith to Lakebase to LTAP: rethinking the database from storage up by Reynold Xin, Databricks Blog, June 30, 2026.

Can a database do two things at the same time, really fast? Can it write one transaction and summarize all transactions at the same time, without bringing the entire system to its knees? That’s the classic data engineering problem: analytic queries have opposite goals from business systems.

This article, written by one of the brightest minds in data processing, explains how Databricks solves this problem in a single database. Now I’m not a PhD like the author, but I’ve seen this problem solved in much easier, less expensive, and perhaps more reliable ways, maybe not in every context, but certainly in a corporate setting. Someone must ask the dumb questions here; might as well be me.

Hey readers! Don’t forget to check out the Spotify playlist at the end of this article. Every Frictionless Decisions article includes a hit 80s New Wave song to help you remember the article’s theme. Let’s jump in!

Subscribe now

You’ve built a good, standard data platform. You’ve responded quickly to requests for reports and data access. Why are the executives in your company still so unsatisfied with their data?

Here’s one possible answer: You’re ignoring their investment decisions.

Data teams don’t understand business decisions as much as they think they do. They understand business systems - ERP, MES, CRM, and many others - and these applications generate huge datasets. Obviously, people need help to make sense of that data, but investment decisions focus on the future, not past business transactions.

Most of the data people use for investment decisions is in people’s minds, not in those systems.

Decision Types

Take a look at all the reporting solutions your data team supports. Group them by the type of decisions they support. Try these three categories.

Operational. Operational reporting focuses on measuring a business process, like shipping products. For example, on-time delivery percentage measures the shipping process, and uptime percentage measures a system’s performance. These reports don’t measure the entire company, only a discrete operation. They’re not strategic.

Performance. Some reports measure overall company performance. An aggregate measure like book-to-bill ratio indicates a growing order backlog; you’ve received more orders than you shipped. They might make an investor feel more confident, but data comes from past activity. They don’t represent a strategic decision.

Investment. Investment data shows where time and money are going toward new business. Engineering development time, for example, indicates the new products or capabilities that your company is prioritizing. Sales opportunities show new customers for existing or new products. These datasets indicate real decisions about the future. They’re not your customer’s or your vendor’s decisions; your company’s employees make them. They show your company’s resources and capacities. They’re inputs to decisions, not just outputs from operations.

Where do you find this data? Investment decisions start at the top of your company. By the time a decision shows up as a business transaction, the strategy is already locked in. Data for the next investment decision is now in the executive’s mind. It’s probably also on their desktop.

Lifecycle Thinking

These decision-data categories mirror “product lifecycle” thinking. Famed Harvard economist Theodore Levitt popularized the model to help leaders think more strategically about their business. He saw that many companies unintentionally followed a one-and-done decision-making process; they created an innovative product, but failed to continue making decisions to extend the product’s life or replace it with the next generation. He saw four phases in the cycle:

• Development: Defining a product and creating demand for it. The goal of this stage is to minimize product launch failures.

• Growth: Successfully launching a new product invites competition. The goal of this stage is to establish brand preference and dominate the market with proactive positioning and pricing decisions.

• Maturity: When a product has fully penetrated the market, strategy shifts to extending its life and defending its market position.

• Decline. Technology changes, consumers get bored, margins shrink, and competition increases. The goal of this stage is to retain customers by proactively moving them to replacement products.

During the growth, maturity, and decline stages, business systems generate operational data. Performance metrics measure these phases. But strategic decisions are made almost entirely before those stages even begin.

That’s why this lifecycle framework is so helpful: it changes a decision-maker’s perspective. They stop focusing on past performance and consider the entire financial trajectory of their products, their broader purpose, and how innovation affects the curve. They ask different questions: Can you predict the lifecycles for a new product? Can you know what stage a product is currently in? Can you use this knowledge to your advantage?

Levitt said, “Leaders know the concept of lifecycle, but don’t know how to use it for advantage…at each stage in a product’s life cycle, each management decision must consider the competitive requirements of the next stage.”

Find the Hidden Data

One of the most common problems I hear from executives is that their IT teams don’t understand the business. Then I hear the opposite story from IT leaders; they think the business people haven’t done enough to explain their requirements. How does this disconnect go on so long and become so painful for a company that they need to call in a consultant?

Executives compare business data to their original plan, and that plan data probably doesn’t live in any of your business systems. Instead, it’s in a spreadsheet on their desktop. It might even be in a PowerPoint slide. But it lines up with the same basic structure as all the business execution data that will eventually show up in your ERP system. Capturing this plan-of-record data in your data platform suddenly creates the missing decision context they are looking for.

Connecting data to decisions doesn’t require you to keep up with the very latest data management technologies. Capturing all the data from your business systems won’t make the connection either. Even scouring the entire corpus of the internet and processing it in real time won’t make a difference. To connect data with decisions, you need to get into the mind of decision-makers and understand what they’re thinking about.

Capture that unseen data, and you’ve closed the context loop for executives. You’ve made the data for all the other phases more useful.

To remind you of this week’s data concept, enjoy Don’t Forget Me When I’m Gone, by Glass Tiger from the Frictionless Data Spotify playlist.

Frictionless Decisions brings you counterintuitive, original, jargon-free news and ideas for connecting data to decisions.

Subscribe now

Almost everyone agrees that faster is always better. Speed might even be the greatest American value. Kids used to walk to school, but now they fly by on electric bikes as I walk to the coffee shop. Parents get real-time alerts when they arrive. We “read” more books faster by listening to them. The mindset affects our corporate data, too. We’re no longer talking about real-time reports, but real-time intelligence.

As a philosophy of life, things can get pretty reckless when going faster is your only goal. Let’s slow down for a moment and ask: does faster data really help your company make better decisions? Is speed really the simple solution for your company’s analytics issues?

Data leaders are confused about the pace of reporting data. One team I know loaded sales order data to their reporting platform on the first Monday of each month. Sometimes that was the next day, other times it was six days after the end of the previous month. People would forget this and constantly ask each other why the data was wrong. At the other end of the spectrum, a corporate data leader posted recently, “Real-time streaming is about to become the most important infrastructure investment an enterprise can make…exciting!”

Good timing is critical for making data understandable. But as a blanket rule for analytics, speed is a terrible strategy. How do you find the right balance?

Matching Speed to Decisions

Here’s an example of matching data timing with decisions. Let’s say your company is launching a new product. Business data for the project comes in three flavors: the original plan, a forecast of results while it’s in progress, and the final results.

Plan of record. The decision to launch a new product is an investment decision. People build a case for the investment, and that data becomes the comparison point going forward. Once you’ve captured that data, it never changes. It’s one-and-done.

Forecast. Once the plan is set in motion, people focus on course corrections. Designers, sales agents, and operations managers see how things are going as they happen. They can’t change what’s already happened, but they can project the future. The sooner that news - their forecast - reaches decision-makers, the better.

Forecasts are usually your smallest dataset, but the most meaningful for real-time visibility.

Actual. Once you’ve shipped the product, that business is “in the books.” Actual results data usually follows monthly financial cycles. This transaction data is your largest data set, and yet it’s the least important for real-time visibility. There’s nothing you can do to change it anyway.

All three datasets have the same shape and form, but they represent different decisions and come from different groups of people. Understanding the nature of the decisions helps you avoid generic solutions like real-time analytics.

Moving data in real-time isn’t always bad, however; it’s a far more reliable way to move data than batch processing, and a lot easier. Operational decisions, like measuring the performance of a machine in real time, don’t require much context. But strategic decisions require lots of context, so data leaders shouldn’t assume that just because they can move data fast, people need to see it in real time.

Be Quick, Don’t Rush

John Mark Comer, in his bestselling book by the same title as this article, says that slowing down isn’t a popular move, so you’ll need to be ruthless in your efforts to avoid the chaos. Many situations you face are rarely as urgent as they seem in the moment. I think he’s right on target; better ideas and reactions usually come when you’ve had time to think about them and consider more factors. You can help people better when you’re not in a hurry.

That’s also true for your data architecture; the best solutions match the speed of data to the decisions people use it for. Be quick, but don’t rush.

To remind you of this week’s data concept, enjoy Dancing in Heaven by Q-Feel, from the Frictionless Data Spotify playlist.

Frictionless Decisions brings you counterintuitive, original, jargon-free news and ideas for connecting data to decisions.

Subscribe now

Have you seen the new Claude commercial? Without explanation, it begins with a short clip of a house burning down. Then the spot delivers its core message: don’t be afraid to ask hard questions. We need this.

That’s the focus of these news roundups: I want to help find ways to avoid the negative effects of data technologies, in your company and in your soul. We need to ask the obvious questions (do AI notetakers really improve communications?) and take them seriously.

Don’t miss the unfolding drama as AI comes after enterprise applications. CIOs, if you didn’t see this war coming, catch up quick. Tableau is also in its sights.

I’ve continued to develop some favorite writers in this month’s list. The articles below suggest a deeper, more thoughtful approach than most. Digesting all this takes a lot of time (which you probably don’t have), so I’ve started condensing and curating the best articles and posts for you.

Give them a read!

AI’s Personal Impact

Sloppy Joe, Anyone? by Clayton Chancey, Praecipio.com, July 15, 2026.

Does anyone read AI-generated meeting summaries? Even if you did, would it be wise to act on them?

Chancey makes a strong case against “workslop,” starting with a definition that will not just make you wary of AI shortcuts but also any work that shifts a cognitive load onto your colleagues. He backhandedly calls out companies for lacking “enterprise metacognition,” where an entire company loses its ability to think about its thinking. Translation: AI can make your company value intellectual laziness. Reader beware.

China’s AI Optimism Isn’t What it Seems by Zilan Qian, China Talk, May 22, 2026.

There’s an irresistible, yet perverse logic behind the AI buildout. This insightful take on China’s “AI optimism” might give energy for your efforts to remain human. Read this and put yourself in the shoes of a normal citizen in a country purely devoted to a technological future.

“Enthusiasm and fear are not mutually exclusive. A person can genuinely believe some AI products are beneficial and feel they have no real choice but to adopt it; can welcome a technology because it seems useful while worried that not mastering the usefulness renders themselves obsolete…China, under this logic, also needs urbanization, industrial upgrades, or AI integration, because history is irreversible and technological progress is inevitable.” Brilliant insight from .

Let Us Now Praise Famous Shape Rotators by Hollis Robbins, Anecdotal Value, July 11, 2026.

suggests that people can shape thought about the purpose of their work even while shaping the mundane things they work on. In the age of AI, we could use more Ben Franklins.

Data engineers certainly fit the role of plumbers. Data flows like water, going anywhere it wants if you don’t have a plan to control it. Polluted data is dangerous. The cleanest data comes from the cloud. If the data is bad, people say it stinks. All the work of a data engineer is done beneath the surface.

Yet the things they shape are entirely weightless. Data engineers manage the memories and plans of an entire company. They shape the way people understand what’s happening in the business. They’re building thought architectures, not just data architectures.

So, to answer the question: Yes, the best data engineers I know are like Franklin. They constantly connect their work to ideas, explaining their work in metaphors. They easily bridge the seen and unseen worlds.

Tech Trends

Up the Stack: How AI’s Escape From the Commodity Trap Risks Enterprise Lock-in by and Akash Kapur, AI As Normal Technology, July 9, 2026

There’s a war brewing between AI frontier companies and corporate software vendors. I’m sure my CIO readers are watching this, but just in case, take note of this story.

An industry falls into the “commodity trap” when its business yields low margins while the users of its products profit, like electricity producers. That’s called the “Bertrand Paradox”: when firms sell heterogeneous products, price competition forces them to sell at the marginal cost of producing. The authors see this phenomenon emerging with frontier AI companies. How will they respond to this new reality?

They’ll move their services up the enterprise software stack. They’ll train on corporate data. They’ll aim for enterprise systems of record. Can AI engines steal SAP’s customers? If fear is an indicator, as this article suggests, the threat is real. If you don’t track this trend, you may find yourself trapped in a new kind of enterprise lock-in.

The AI “5-Layer Cake”: How Nvidia’s CEO Explains the Real Economics of Intelligence by Shane Collins, Activated Thinker, March 15, 2026.

Need help understanding the economics and transformation of AI? A great acrostic can frame all your work in its larger context. For example, I learned networking fundamentals by remembering “Please Do Not Throw Sausage Pizza Away.” Jensen Huang has a new image for you: the five-layer cake (ECIMA). It’s time to make up your own acrostic to remember this.

The Tableau Exodus Has Begun by , Super Data Blog, June 4, 2026.

I highlight this essay simply because I love the title. Why is anyone surprised that “game-changing” software platforms can fall out of vogue so easily? Well, we tried solving management problems with software. Companies that did this will swap out Tableau for Claude in a heartbeat.

If you need more evidence of the swap, check out these posts coming out of the latest Tableau conference: “Extend or Replace?”; “How Flexibility Becomes Debt”; “Is Tableau Still Valuable?”; “Tableau Isn’t a Black Box. But It’s Something Almost as Bad.”

AI solutions are exposing a management gap: software tools don’t solve integration problems. It takes a deep understanding of processes, systems, data, and people to do that. Can Claude do it better than Tableau? Probably.

Data Engineering

The Forgotten Measure of Data Quality: Decision Quality by , Modern Data 101, July 13, 2026.

Every data leader knows the purpose of their work is to connect data with decisions. So why do they spend all their time on technical solutions? Wernicke suggests we should honestly ask if our solutions are helping a company make better decisions.

Then he goes one step further, which I love: evaluate the decision framework of your company. Are the decision flows aligned across the organization? Do people in different roles receive the data at the same time? Is the way they understand the data consistent? Does the adoption (use frequency) of your tools reflect this?

If you like thinking about how a company thinks, you’ll love this smart piece.

Your Data Warehouse Isn’t Integrated Just Because the Tables Are in One Place by Ben Rogojan, Seattle Data Guy, July 18, 2026.

Fix data at the source; that’s a core concept of data architecture. “The best managed companies, from a data perspective, integrate data at the application layer, not the data warehouse,” says . Sounds like one of the “00001010 Commandments of Data.”

Thesaurus by Jessica Talisman, Intentional Arrangement, July 19, 2026.

I met a data engineer at a coffee shop. When I told him I’d written a book about data strategies, he asked me for one piece of advice he could use. I said, “You’re a librarian, not a technologist. You organize data. Go make it easy for people to find what they’re looking for.”

You can do that too. If you want to develop your own understanding of the data framework underlying large language models, here’s a great place to start. takes a library science concept and translates it into a data structure. You don’t need to become an expert or even a practitioner to appreciate taxonomies.

How to become a Data AND AI Engineer by Efran Hasami and Alejandro Aboy, Pipeline to Insights, July 15, 2026.

Enjoy a fun conversation between two of my favorite data engineers, and , writing on Substack. Solid, enterprise thinking here.

To remind you of this week’s data concept, enjoy Burning Down the House by Talking Heads, from the Frictionless Data Spotify playlist.

For the full story about making data flow faster and better, check out Frictionless Data on Amazon.

Frictionless Decisions brings you counterintuitive, original, jargon-free ideas for connecting data to decisions.

Subscribe now

You might think that administrating people’s access to data is the most boring part of working in IT, but it’s one of the coolest parts of the job for me.

Business people usually approach security reviews as a necessary evil, a list of items on a policy compliance checklist. That’s not the case with a good data strategy. When you sit down with a business leader and ask who should see their data, it’s like you’ve given them a gift. You’ve recognized their authority and told them that they can hide data from anyone, at any level of detail they decide, making their decisions effective everywhere.

You’ve given them control. After all, it’s their data.

Flip the Script

CIOs know low-level service tasks can improve or destroy an IT department’s reputation; people grade you by their experience when they contact the help desk. Managing data security falls squarely in that category. You can flip the script on this narrative and use data security to build partnerships with leaders. You can transform people’s daily lives by giving them exactly the data they need to do their jobs.

Imagine a world where nobody complains about not having enough access, and nobody worries about exposing more data than they should. That’s the promise of a good analytics security solution.

Companies often assign technical people the responsibility for managing internal data security. Think about it. Do IT people understand other employees’ jobs well enough to know what data they need? Of course not. Do they know which data is sensitive or benign? No. IT people just see files on a server. When you put IT people in charge of securing internal data, they just ask people what data they need (through far too many emails) and give it to them.

Who Knows the Data?

The most granular way to secure business data is called “cell-level” security: controlling data access at the intersection of the rows and columns within your database tables. To do that, you need to know your data, who owns decisions about it, and who needs it for their job. IT people aren’t experts in either area; you need to find business owners to make these decisions.

You need two people involved in every access decision, one who knows the data (data owner) and another who knows the people (role owner):

Data Owners. These people know the meaning of the data (like pricing or costs) and have management authority to decide which types of job roles need to see that data. For example, the VP of Marketing (Helly R., in my illustration below) owns all pricing data. As the data owner, she decides if financial analysts, sales analysts, product engineers, and HR staff (for example) need access to pricing data. She isn’t concerned about which individuals get access to pricing data; she only cares about which job roles need the data.

Role Owners. These people know who is qualified to work in a given job, and they usually manage people working in that role. For example, the VP of Product Engineering (Mark S., in my illustration below) owns the “product engineer” role used to assign access to data. When someone (like a new employee) wants access to data, he approves their request to become a member of that role. If he thinks product engineers need pricing data to do their jobs, he can discuss it with Helly on behalf of all product engineers.

Here’s an example of how these permissions look in a table:

This matrix demonstrates a fundamental feature of scalable, granular, understandable security: it relies on logically structured master data.

It also relies on an effective “identity management” solution, like Windows Integrated Security. People should be free to use whatever tool they want to visualize, pivot, or simply report from the data. That’s a core tenet of a frictionless data strategy: pushing security enforcement to the lowest level of the system makes your data “tool agnostic” for end users.

Beyond Compliance

Giving people control of security decisions builds a partnership that goes far beyond managing permissions.

Two dozen executives from all areas of the company - finance, sales, manufacturing, marketing, and even R&D - welcomed me into their offices every quarter to review the permissions. We’d usually spend the first ten minutes of that hour reviewing people and their permissions, identifying a few updates along the way. Then we’d spend the rest of the hour discussing upcoming plans for the data warehouse. It was a great time to ask them what new data they could use for their team’s decisions.

Often, those discussions turned into brainstorming new project ideas, using whiteboards to sketch out relationships between data sets, and thinking together about what new insight they might inspire.

Just saying that the “business owns the data” isn’t enough to convince people to trust you. If you invest in these administrative processes and system features, you show them that you really believe they own the data.

To remind you of this week’s data concept, enjoy Career Opportunities by The Clash from the Frictionless Data Spotify playlist.

For the full story about making data flow faster and better, check out Frictionless Data on Amazon.

Frictionless Decisions brings you counterintuitive, original, jargon-free ideas for connecting data to decisions.

Subscribe now

Spending too much time thinking about AI can fool you into pretending you’re a sophist, a paid teacher of philosophy. Don’t worry, I learned that lesson long ago.

But I do want to help find ways to avoid the negative effects of data technologies, in your company and in your soul. This month’s news roundup articles focus on AI’s personal impact, global tech trends, and data engineering. You might notice how I’ve developed some favorite writers in this month’s list. Oh, and have fun with the robots!

The articles below suggest a deeper, more thoughtful approach than most. Digesting all this takes a lot of time (which you probably don’t have), so I’ve started condensing and curating the best articles and posts for you.

Give them a read!

Amor Fati Entropy by Hollis Robbins, Anecdotal Value, July 6, 2026.

is a writer’s writer. This little piece on the redundancy of Internet-written content reminds me of Herman Melville writing about the ship’s ropes while chasing the White Whale. Just read it and enjoy.

I asked Robbins if she had some advice, simple guardrails, for preventing AI from infiltrating my writing? She says, “Focus on things only you know that it doesn’t know about! And push back hard (the thumbs-down button at the end of every chat is there for a reason).”

How to Plagiarize - by Jessica Talisman, Intentional Arrangement, July 6, 2026.

A weighty essay here. gives us the nuance we need to personally resist.

As a writer, plagiarism terrifies me. Not that I think anyone would want to steal any of my ideas, but that I might accidentally steal from someone else. A simple rule helps me: never write with AI. Never. Never paste any text from the internet (or any AI tool) into my writing. This doesn’t solve the problem but works as a starting guardrail. Another guardrail: always include attribution as an editing pass.

Plagiarism is a massive problem, and almost everyone who uses AI for writing is guilty, even if unknowingly. You can even plagiarize your own writing, like I just did in the previous paragraph.

The Robots Are Here by Jordan Schneider, China Talk, July 6, 2026.

Don’t discount robot dogs. and his guests draw a very helpful analogy to DJI creating the market for drones. Five years from now, you might wish you’d read this.

What Actually Happens When a New Chip Powers On for the First Time by Dr. Anu Asokan, LinkedIn, June 27, 2026.

When your confidence exceeds your knowledge, you become an armchair quarterback. That’s me when it comes to launching a new chip. In this fun, short LinkedIn post, the author gives everyone, even those not in the business, a window into the lived experience of design engineers.

Adopting AI Augmented Knowledge Work by Clayton Chancey, Foray Consulting, June 4, 2026.

I’ve recently started following Chancey. He’s really thoughtful about using AI as a tool, and he brings the credibility of an integrator. In this video presentation, he explains how AI is best used to accelerate processes you already understand. Makes a lot of sense to me; I hear the same message from most real experts and I’ve developed the same perspective in using AI for writing.

Chancey develops a solid thesis here: think about AIs impact on everyday workflows. This prepares you best for the transformation.

Forward Deployed Engineering Is About To Get Diluted by Ben Rogojan, Seattle Data Guy, July 6, 2026.

I love ’s data engineering updates; jargon-free, friendly, perceptive, and written with a ton of practical credibility. I think we’re all going to be FDEs five years from now.

How To Transfer Knowledge Every Time A New AI Buzzword Gets To Your Feed by Alejandro Aboy, CQL (Career Query Lessons), June 7, 2026

I’ve honed in on a couple bright, young writers for data engineering insights and to complement my own focus on connecting data to decisions. is one of the best.

“Using old terminology makes you sound outdated even when you’re describing identical work, so please don’t use terms from 2 weeks ago, which is the pace in which AI tech deprecates definitions,” he says. Love it.

I always felt that most data architecture questions boil down to “Where do we put the logic?” It’s a great way to cut through jargon. Non-technical nuggets like this can help you find your way around the ever-changing terminology world.

How Anthropic enables self-service data analytics with Claude by Chen Chang et al., Claude.com, June 3, 2026.

Check out my recent post reflecting this Anthropic article. I call it “Anthropic’s Common Sense Approach” to using AI in analytics.

I wouldn’t worry about Claude eliminating data analysts. Most of the article sounds like something I would write in Frictionless Decisions, in the sense that it emphasizes integration at the data foundation level. We can take this “data foundation” idea much further, even to building the foundation within business systems themselves. Maybe a future Frictionless Decisions article?

To remind you of this week’s data concept, enjoy Cruel Summer by Bananarama, from the Frictionless Data Spotify playlist.

For the full story about making data flow faster and better, check out Frictionless Data on Amazon.

Frictionless Decisions brings you counterintuitive, original, jargon-free ideas for connecting data to decisions.

Share Frictionless Decisions

Most everything you read tries to convince you that data will help your company make better, faster decisions - if you just use the right approach and the latest technologies. But more information and more layers of technology usually lead to more chaos, not less. Software alone cannot remove the friction from your data; how you manage the “flow” of data into decisions has a greater influence on how your company thinks.

Let’s think about how your company thinks, consider how data feeds into your company’s everyday decisions, and look at the people side of data and how an entire company uses it. Be a thought architect, not just a data architect.

Traditional Decision Flow

In most companies, data flows through people, from analysts to managers to executives. Data gets compiled – audited, summarized, modified, augmented, filtered, and enhanced – at every level, by many people and many departments. Analysts draw conclusions and steer the data toward their point of view. Layers of human and system interpretation get baked into every step along the way to the executive, leaving the data’s message softened and disconnected.

When analysts manage and model data in isolation, each analyst has the chance to add value or confusion to the decision flow. Many feel intuitively that knowledge work should not be governed at all, but while it might seem that this diversity helps uncover insights, many who experience it describe it as the “wild west.”

In my experience, most analysts would rather spend their time making business decisions than managing data. When you, the data leader, manage the flow of data into their decision processes, you’re not governing their decisions. You’re freeing them to do what they do best: evaluate the data and make recommendations to improve the result.

Would you expect the message (the story the data tells) to get clearer when it flows through people? Will bad news easily flow up to the top of an organization? Maybe, but the phenomenon of bad news getting better as it gets communicated upward has an official name – the “Mum Effect” – given by the American Psychological Association.

Building Backwards

Henry Ford gave us the classic example of reversing the direction of workflows. Ford and his engineers did not invent the car; they only applied the idea of assembly lines to building cars. Before the Model T, car building was an “asynchronous” workflow: a single process, but none of the steps happened in any particular order. Specialized workers found the parts they needed, went to each car and performed their craft, then went to the next car and did it again. The car was finished when each worker independently completed their part.

Ford finally realized that the overall approach was backwards. This required a lot more thought about the steps involved in building a car. Once they got the flow of work figured out, it was enormously powerful. By 1925, Ford was building about 10,000 Model Ts every day.

Ford’s challenge was getting the right automobile parts to the right people at the right time, at the right level of finish. But for us, it’s about delivering the right information to the right people at the right time, at the right level of finish.

Bypassing Interpretation

Look closely, and you’ll start to see how corporate decisions repeat; they follow a sequence, a predictable flow.

For example, a semiconductor company makes hundreds of decisions every quarter about which products to launch, but every one of those decisions is a return-on-investment (ROI) decision. The data is different for every product, but the structure of the decision is always the same. Every ROI decision requires the revenue amount, manufacturing costs, and engineering costs for product design. Business managers need each of these data points when deciding which products to launch and when looking back to assess their decisions.

Just imagine a company where this data flows to every decision-maker concurrently: the CEO receives data in their inbox at the same time the analysts receive updates in their analytics tools. I want you to see the simple concept here: in the “frictionless” solution, data bypasses interpretation. Decision-makers don’t hunt for this data, and they don’t need to reinvent the process for aggregating it to the level of their decisions. Frictionless data doesn’t try to make existing data flow more efficiently; instead, it makes data flow without mediation. It flows to people, not through people.

Turning the Tide

Reversing the flow of data in your company might sound good in theory, but how do you turn it around? There’s no silver bullet; you achieve this scale by centralizing business logic and delivering data to the right people at the right time and with the right detail. You establish trust through processes that measure data quality, enforce accuracy, and link decision inputs to master data.

Start with some simple, high-level indicators that can help you see where to begin:

• Do top decision makers receive data from the system or from people? A high-velocity data solution delivers fully qualified data to decision makers, at every level, every day, all at the same time.

• Does core master data look the same across all your systems? Business processes always share master data (like products or customers) even if the systems don’t make it easier. How well you’ve synchronized this master data across your company gives a clear measure of flow.

• How long does it take for an organization change to get reflected in all reports? Organizational changes are one of the most common challenges of a data solution. It’s the true test of data agility. A high-velocity data solution updates every report available to users the same day that the organization structure change gets made.

• Do you know the number of reports your company publishes? The more reports you have, the less your data is flowing naturally into decisions.

• Is it difficult for a user to find detailed, matching data for a key metric? A good data strategy makes it easy to navigate from a metric to the business transaction or forecast behind the metric. When people need to navigate to a different system or a different report to see the details behind a metric, it’s not flowing easily.

Your company’s decision-making capability is an asset; data flows into it, and the results complete the feedback loop. If people in your company feel frustrated with their ability to use data to make decisions, you need to reverse the flow.

To remind you of this week’s data concept, enjoy Once In A Lifetime by Talking Heads, from the Frictionless Data Spotify playlist.

For the full story about making data flow faster and better, check out Frictionless Data on Amazon.

Frictionless Decisions brings you counterintuitive, original, jargon-free ideas for connecting data to decisions.

Share Frictionless Decisions

A year ago, I wrote a simple post about data warehousing, called “Always Warehouse the Data First.” Honestly, I wondered if anyone would think this decades-old data management approach sounded relevant. People online weren’t just asking if data warehousing was relevant; they were actively predicting how AI would eliminate data management. I’m sure that got a lot of clicks, but I didn’t feel qualified to step into the debate.

Imposter Syndrome strikes back.

Making matters worse for me, many experts I follow have completely abandoned data warehousing in favor of AI solutions for organizing data. A few weeks ago, for example, Modern Data 101 published a piece explaining how AI can resolve differences in business definitions using a good ontology. The author asks how you can get AI to consistently deliver a correct answer to the question, “What was Q3 revenue?” to everyone in a company. Their solution is very technical, but I’ll summarize it as ingesting all the language used across different departments and mapping it to a single, larger semantic model.

I never had a problem delivering accurate financial totals to everyone using data in my company. We did this with good data governance practices in a data warehouse. I love ontologies (thank you, !), but I wondered why anyone would ever think you needed AI to solve this problem. It seemed like the classic “Law of the Instrument“ bias: when you only know how to use a hammer, everything looks like a nail.

Claude Chimes In

To my surprise, Claude chimed in on this problem a few weeks ago. Anthropic’s own data analytics team addressed the relationship between Artificial Intelligence and Business Intelligence. They address the age-old question of how a data team can balance self-service analytics with the need for trusted, accurate results. It’s a brilliant piece.

You’ll find two core management insights in their article. First, business analytic questions are different from other types of knowledge; there’s only one correct answer. You’re not looking to resolve ambiguity. This table explains the differences they see:

Second, while AI is amazing at helping non-technical people navigate data, it will only be successful in business analytics when it is querying a well-governed data foundation. A few key quotes from their article:

• “Standard data engineering and data quality practices such as dimensional modeling, shift-left testing, freshness and completeness checks on critical pipelines all still apply (and we won’t relitigate these).”

• “The data foundations layer is aimed primarily at ambiguity: if revenue, for example, resolves to one governed dataset instead of forty plausible candidates, the problem largely disappears before the agent ever has to search. It’s also where the first staleness defense lives, since the same repo that defines the canonical models is the natural place to enforce that they stay current.”

• “The goal is that when an agent searches for a concept, it finds a single governed answer.”

Common Sense Prevails

My friend Yakov Shkolnikov, one of the smartest AI engineers I know, has been making this case for years. When I shared the Claude article with him, he summarized it by saying,

I’m happy to report that common sense has prevailed. Or at least it has a strong foothold. Thank you to Frictionless Decisions reader Diana Nekhorosheva for sharing the Claude essay with me!

If you’re struggling to get your AI tool (or any BI tool) to deliver accurate, trusted answers, don’t forget the Seventh Commandment of Data: Always Warehouse the Data First.

To remind you of this week’s data concept, enjoy Party Out of Bounds, by the B-52s, from the Frictionless Data Spotify playlist. When you listen to this song, don’t forget that it’s a metaphor (like most B-52s songs)!

For the full story about making data flow faster and better, check out Frictionless Data on Amazon.

Frictionless Decisions brings you counterintuitive, original, jargon-free ideas for connecting data to decisions. This article is part of my series, The 00001010 Commandments of Data.

Share Frictionless Decisions

All the buzz around analytics today might have you convinced that there’s gold buried somewhere in your data. As Billy Beane famously said in Moneyball (the story that started the buzz), “We are card counters at the blackjack table. And we're gonna turn the odds on the casino.” But everyone saw the movie, so it’s really a race and you’re worried you might fall behind.

I’ve got news: if you think analytics is mainly about finding "hidden insights," you're wrong.

Business analysts spend a lot more time explaining what happened and why the results don’t match people's expectations rather than finding brilliant ideas. Questions like “We had 1,000 students last year, now we have 1,100. What changed?” or “We booked some big orders this quarter, but our backlog only went up by $100k. What happened?” consume most of their time.

I’ve always used this “Sixth Commandment of Data” to help analysts see internal integrity in the numbers. By measuring the grand totals before measuring the different parts of a set of data, you can ensure that the beginning and end points of the numbers agree with (“reconcile”) all the changes in between.

Defining "The Whole"

There's no way to explain what changed in your business unless you know how much business you had at the beginning and what you now have at the end. That's what I mean by measuring the "whole": knowing the grand totals of your most important data sets.

• For a manufacturing company, it's your customers. What's the value of all the orders on the backlog?

• For a university, it's your students. What's the total number of enrolled students?

• For a SaaS software company, it's your subscribers. What's the value of their active monthly subscriptions?

• If you're managing a warehouse, it's your inventory. What's the value of all the items in stock?

• For any company, it's your people. Measure the total count of employees (headcount).

I call this the "book of business": knowing the total amount (value and quantity) of all the assets you manage. Most analytic questions try to explain what happened between the beginning and ending values of these totals. You can apply this approach to any set of data for your company; the more broadly you define the “whole”, the more questions analysts can answer with the data.

Measuring the Change

When you measure the whole book of business, you can help analysts answer questions about what changed. You can calculate the amount of the overall change before measuring all the different reasons the numbers changed.

Let's say your organization is a university, and you started the year with 1,000 students and ended with 1,100. Take these steps to measure the whole change:

• First, calculate the total change: the difference between the total student count at the beginning (1,000) and what you have at the end (in my example, an overall increase of 100).

• Once you know the overall change in the data, you can start to measure the parts and calculate easily defined changes that everyone understands. You added 267 students through new enrollments but lost 128 students through graduation. Everyone agrees on how to measure those numbers.

• But that doesn't explain all the changes; you've got an additional, unexplained drop of 39 students. Maybe that was students who unexpectedly dropped out?

By measuring the whole before measuring the parts, you've "cornered" the unknown changes; you haven’t explained all the changes until all the explanations equal the total change amount.

What I’m showing here is just the normal way analysts think. But embedding that thought process into all the data is how a great data strategy accelerates their work. What's the best explanation for the unknowns? Ask the analysts. Measure the changes like this and they'll figure out what's really happening in the business.

Different industries use different names for the activities that normally change their book of business. In manufacturing, it's "bookings". In SaaS software, it's "churn". In retail, it's customer “retention”. How you measure the "whole" depends on your business model, and my simple examples might not describe your industry; a manufacturer can easily measure customer value from hard orders in their system, but measuring customer value for software subscriptions requires more estimation. In any case, following the definitions your company uses to report the financial results usually makes sense.

Defining this "whole" for your company sets the cornerstone for most analytics, and you can apply this approach to almost any type of data. That's why I'm giving you this sixth commandment:

Always measure the whole before measuring the parts.

To remind you of this week’s data concept, enjoy Head over Heels, by The GoGo’s, from the Frictionless Data Spotify playlist.

For the full story about making data flow faster and better, check out Frictionless Data on Amazon.

I’m not an AI architect, not a subject matter expert, and I haven’t personally managed any corporate AI implementations. In other words, I’m probably just like you.

I do, however, hear from you, my readers. You’re making technical decisions at a faster rate and higher cost than I’ve seen before. Although the AI tools you’re using are all similar to each other, your approaches to managing them vary widely. I want to help you avoid repeating the mistakes I experienced when previous “game-changing” analytic solutions were quickly adopted.

Let’s talk about governance. Even with the best tools, the process you follow for managing them ultimately determines how well your company uses them to make decisions.

It’s all about balance: giving analysts autonomy without losing the benefits of control and coordination. Analysts provide the most value when they can freely explore data, but you can’t synchronize decisions with a data free-for-all. Without hands-on decisions about change control, even the best strategy can fail.

Those of you I’ve talked with are taking three different approaches to governing AI in the business intelligence (BI) space:

• Decentralized access. With full access to all data sources, analysts independently develop data models, generate analysis, draw conclusions, and report to management. AI accelerates reporting by eliminating technical barriers to accessing and interpreting data.

• New software layer. Third-party software provides tools for managing data definitions (semantics) and provides direct access without losing the governance needed for alignment. They may include knowledge mapping capabilities, a critical feature for AI language processing.

• Data products. An interpretive layer provides access to enterprise application data using the standardized MCP (model context protocol) design. This leverages native source system security without adding new processes. You can almost think of all data architecture as building context for decisions.

The first approach - decentralized access - mirrors the “democratic” approach advocated by many software vendors over the years. Tableau Software, for example, appealed to users for this reason: every user gets a license and the freedom to build their own data models. The problem with this approach isn’t the software - interactive visualization truly changed the way people interact with data. But most companies now recognize the chaos that thousands of dashboards create for their overall decision processes.

The second approach - adding new governance software - is a step in the right direction. Unfortunately, it tries to solve the complexity of governing your data by adding more complexity. That might be okay if you don’t have a data team to handle the more sophisticated solutions, but your company’s data models are your own secret sauce. One-size-fits-all solutions won’t save you money and will limit your capabilities.

The third approach - governed data products - most closely aligns with the Frictionless Data solutions I managed and write about.

Federal Framework

Governance is the framework for making decisions about your data platform. In this article, I compare the democratic, centralized, and federal data governance models. The trend of shifting work from traditional IT teams to those who understand data modeling best - the analysts - isn’t new. Tableau and Power BI, for example, introduced this model over a decade ago.

Without some kind of centralized process for qualifying content and solutions, this shift can lead to chaos, just like it did with every previous BI transformation.

Governance can eliminate the chaos and create leverage. The goals of governance include:

• Alignment - ensuring consistent, trusted answers from data across functional teams, management layers, and time periods.

• Scale - delivering data to everyone at the same time, without processing or interpretive delays.

• Cost - eliminating duplicate work, processing, and missed value-add opportunities.

Governance should be formalized early on in your journey. Keep in mind three considerations as you ramp up your solution:

Security

This is the hidden scaler for decisions. Security makes certain that people have complete access to only the data they need to do their job, without adding cost and complexity. It cannot be an afterthought; defining permissions and the ability to enforce them must be designed into the solution’s foundation. MCPs and 3rd party solutions probably don’t know your data well enough to deliver this value out of the box.

Make a clear distinction between operational security and analytic security. People view data in two directions. Operational teams view data through the lens of a business process, like order fulfillment. Business management teams view data vertically, across all functions; for example, they view the business through the lens of a P&L. These views imply different security frameworks.

Source Data

It’s easy to get so excited about the data modeling capabilities in AI that you ignore your biggest problem: the data you need the most is the data you don’t have in any platform. AI closes the data access gap, but better source data is equally important. Take a broad view of the source data you need for decisions:

• Enterprise systems. MCPs accelerate access to most of your operational data, replacing traditional ETL processes. Ask if AI outperforms the industrial-class solutions your data team has (hopefully) perfected over the years. If the answer is “yes,” use AI to transform that process.

• New product data. Most of this data does not exist in your enterprise platforms. You probably won’t find off-the-shelf MCPs for this type of data. It’s also your smallest data set, so stay aware of the proverbial case of using a sledgehammer to kill an ant.

• Extended data. Some of your most important data is materialized through logical shortcuts, such as end-market data. The processing and time required to link demand-fulfillment flows to end-customer decisions are beyond the reach of traditional business systems and analytics platforms. AI can solve this.

• Business Intelligence hierarchies. One of the most important values you get from a BI solution is aggregating data into the categories executives use for their communications. The best BI solutions recognize this and manage it in the most agile, responsive place they know: their data platform. These categories will overlap with the taxonomies used by your new AI language models.

• Narrative. The semantic descriptions behind variance explanations are data; it’s also proprietary to your company. Your AI platform should give you the ability to explicitly define concept relationships instead of leaving them to system-generated probabilities. Controlling these taxonomies adds efficiency and alignment. This is likely a new techno-functional skill your data team needs.

• Data quality. Managing and automating data quality requires a solution for defining business rules and closing the loop between reporting data and business systems.

Synchronizing Financials

Financial data links business processes with strategic decision-making. Ensure these connections are maintained throughout the platform. Without maintaining these connections, any analysis that uses financial data to measure the business will not flow naturally from your solutions.

Correcting Claude

I’m just as excited about using AI as you. Claude Pro just automated my data modeling work for Common Ground, the non-profit we support. It pulls the demographic data together for our annual report in about a minute. Even developing the program went really fast; together, we built the model in just a few hours. It scales up, too. I can use the same model to analyze any city in the country.

I corrected Claude’s work dozens of times as we built it; other people call this “training.” But Claude did the heavy lifting and educated me about the source data along the way. I have no interest in manually developing automated (API) connections to the census data. Researching the data relationships and geospatial tables doesn’t interest me either. AI gets all credit for this solution; I’m happy to hand over the programming work to AI.

But AI won’t govern itself. That’s the core wisdom I think you need as you step into using it.

To remind you of this week’s data concept, enjoy Secret Separation by The Fixx from the Frictionless Data Spotify playlist.

For the full story about making data flow faster and better, check out Frictionless Data on Amazon.

One year ago today (!), Frictionless Data was published. Whew! Thank you for reading!

I’ve spent three years now interrogating the question, “What’s the secret for connecting data to business decisions?” Writing on this subject led me to explore not just my own solutions, but also the work of thought leaders past and present. The weekly newsletters helped unpack additional innovations, like my Financial Master Data (FMD) series, the 00001010 Commandments of Data, and Shaping a Data Team. Thank you for letting me explore and develop these ideas with you.

I’d especially like to thank everyone who supported my work: pre-readers Bruce Kiddoo, Walter Curd, and Morgan Watts; creators Eric Cummings, Michael Cummings, and Mike Koelsch; Business Expert Press; TruGlobal; endorsers Bruce Kiddoo, Nishit Sahay and Ibrahim Shariff, and for recommending my newsletter.

Summer Writing

I’m taking the summer off from weekly newsletters! New, longer-form, client-specific material has been in the works for a while now; I’ll publish any ideas that emerge from that if they seem helpful to all my subscribers. You’ll continue to receive news roundups, book reviews, and a few rants from me throughout the summer, I’m sure. Also look for a webinar sometime this summer, focused on semiconductor solutions.

This is your chance to unsubscribe :)

Summer Listening

I hope you’ve had as much fun listening to 80’s hit songs as I’ve had researching them. They really do connect to these ideas about data strategy! Here are some favorites so far this year:

1. Secret Separation by The Fixx

2. Songs From the Edge of the World by Siouxie and the Banshees

3. Where’s My Mind? by The Pixies

4. See the Lights by Simple Minds

5. The Sun Always Shines on T.V. by a-ha

Summer Reading

Looking for a good book to read? Here are some of my favorites so far this year:

Against the Machine by Paul Kingsnorth

If the movie The Matrix were a non-fiction book, this would be it. You won’t find a writer thinking further outside the box than this Kingsnorth. It’s a mind-bender, a page-turner, and you will not come away unchanged.

Fahrenheit 451 by Ray Bradbury

Bradbury might have written this book yesterday. Get out your torches. Just one update needed: not only has reading disappeared but also writing. How did that happen? Well, all we care about now is speed.

Working by Robert Caro

Caro, probably America’s greatest biographer, invites you into the writer’s life. Yes, writing is work. This book inspires readers to understand their work as a calling, a craft, and…working.

Thinking Fast and Slow by Daniel Kahneman

Those who’ve followed my writing for long might see how I gravitate toward the cognitive sciences. Here’s a classic. Find out the difference between the intuitive, automatic thought processes that help you make quick decisions (and repeat your biases) and those slower, contemplative thought processes where you evaluate and develop beliefs.

Chip War by Chris Miller

Do you love history, technology, economics, and…data? Throw in the epic story of the invention of semiconductors, and you’ve got a winner. You’ll see my review of this book later this summer.

Substack Recommendations

Here are the Substacks I read! You might enjoy them too.

CQL (Career Query Lessons) | Alejandro Aboy

Ale writes from the perspective of an AI and data engineer. He’s got great architectural vision, and if you’re interested in this space, he’ll help you think beyond the fear of falling behind. It’s a calming voice.

Intentional Arrangement | Jessica Talisman, MLS

I think I’ve learned everything I need to know about the fundamentals of AI knowledge management from Jessica’s work. Think of it like the Thomas Guide for AI.

The Incomplete Guide to AI | Yakov Shkolnikov

Yakov always thinks outside the box, and unlike me, he knows everything about the inside of the box. Fun reads!

ChinaTalk | Jordan Schneider

There’s a mountain of content here, so I focus on the semiconductor geopolitical posts. Chris Miller (author of Chip Wars) is a regular contributor.

The Abbey of Misrule | Paul Kingsnorth

Everyone working on the inside of the Machine should keep an eye on the motives of the Machine. Kingsnorth writes songs from the edge of the world, so to speak. If someone wrote a Substack about their journey to throw the Ring into the volcano, it would sound something like this.

AS ALWAYS, I’d love to hear from you, even if it’s just to introduce yourself or catch up. Understanding your business intelligence journey helps me help you. Reach out anytime via LinkedIn. Thanks! ZH

For the full story about making data flow faster and better, check out Frictionless Data on Amazon.

To celebrate the one-year anniversary of publishing Frictionless Data, I’m sharing the article I used to announce it. Enjoy!

Subscribe now

Way back in 2011, the film Moneyball told the story about how the Oakland A’s used data to outsmart the old-school scouts and coaches. The team’s General Manager, Billy Beane (played by Brad Pitt), revolutionized baseball with data-driven decision-making. The small-market, low-budget A’s couldn’t afford the salaries other teams would pay players in free agency, so they had to find new, less expensive ways to win.

For example, Beane ignored his scouts (who relied on their personal experience and insight) and found washed-up hitters who got on base a lot. The scouts pointed out that all these players had issues, like injuries, questionable behavior, and aging arms. But that’s what made them attractive to Beane. It made them affordable.

The movie came out in 2011, after American companies embraced analytics as a core business strategy. How Beane’s data-driven baseball decisions would play out in the business world wasn't clear when Michael Lewis published Moneyball (the book) in 2003. It was the movie that cemented the idea of using data for decisions in popular culture. Business leaders already believed that idea by 2011, and even today, they use the term “Moneyball” as a shorthand for using data to get ahead.

Today, everyone agrees that companies should use data for all decisions. They’re just not sure how it really works.

The CIO dilemma

The real Billy Beane was a business manager, just like you. Management chronically underfunded and understaffed his team, but still expected them to compete with top teams like the Yankees and their seemingly unlimited payroll. The Oakland A's won 102 games that season but failed to win a playoff series, then the rest of the league quickly caught up with their data-driven ideas.

The A's finally left Oakland last season. They stopped gambling and headed to Vegas, where they'll try to attract a bigger fan base and (presumably) compete by spending money, just like everyone else.

That story illustrates the challenge facing every business leader today: can you close the gap between data and decisions by spending money on AI? Most companies do the same thing the A’s finally did, fixing their problems with expensive software solutions.

The problem with data is not about capturing more of it. It’s about connecting data to business decisions.

A Management Approach

I couldn’t find any books focusing on how to make this connection, so I wrote one. It’s called F(r)ictionless Data: Solutions for Better, Faster Decisions. It’s not a technical read – far from it. I’ve purposely written it with entertaining, real-life anecdotes from my 25-year career managing corporate data teams.

I’m excited to announce that Frictionless Data is available on Amazon today.

In Frictionless Data, I explain how to make the connection in four parts:

• See the Friction. Data isn’t helping your company make decisions, so I’ll help you see what needs to change.

• Reverse the Flow. Your data flows in the wrong direction – that’s the problem. Before we fix it, I’ll help you see the “why” behind all the changes we’ll make.

• Build the Framework. To make this change, you need a deep understanding of data's shape and flow. That’s how we’ll connect the dots between architecture and analysts' thinking.

• Influence the Business. Data should move with every business decision. That means you’ll need to change the way you work. I’ll show you specific management practices and policies that make your data trusted and free-flowing.

I believe you can do this at your company. I’m here for you. There’s no silver bullet for better decisions, but managing your data better, like I’ll show you, is your best hope to make the connection. Let’s game the casino together, not like the movie, but in real life. You don’t need new technologies to do this; you need new management skills.

There’s a better way: Frictionless Data.

Comments from pre-release readers:

“I like the short, punchy chapters, your inviting voice as a narrator, and the clear, actionable takeaways.”

“It addresses a very important, but understudied, key area behind a firm’s successful business intelligence strategy. The book is written in an easy-to-read conversational style.”

“You've got a great knack for making an otherwise dry subject into a compelling read!”

“It definitely hit the right notes!”

“Battle-tested strategies through real-world case studies.”

“It not only met but exceeded my expectations.”

“I think it will appeal to a broad range of information professionals, from managers to the C-suite, and it successfully communicates its ideas and strategies in a refreshingly jargon-free way.”

“The macro-level perspective on data integration is interesting.”

To remind you of this week’s data concept, enjoy The Card Cheat, by The Clash, from the Frictionless Data Spotify playlist.

For the full story about making data flow faster and better, check out Frictionless Data on Amazon.

Frictionless Decisions brings you counterintuitive, original, jargon-free ideas for connecting data to decisions. Every week.

Share

In his recent study, How to Create an Optimal Data & Analytics Organizational Model, Gartner analyst Jorg Heizenberg said you need 27 different jobs to build the ideal data team. But at the same time, most companies never measure up to that standard. Many combine all 27 capabilities into a small team of senior data engineers.

At worst, they look for all 27 capabilities in one person.

Just look at the Chief Data Officer job itself. Companies started adding CDOs to their corporate boardrooms about twenty years ago, tacitly acknowledging that CIOs weren’t translating data into better decisions. But today, the average tenure for a CDO is about 2.5 years, among the lowest of all C-suite staff.

With all this confusion, how do you know where to begin when forming a new data team?

Four Leadership Domains

I start by appreciating the word “team.” That will help you avoid building a monolithic technical team and move forward without thinking you need to hire another person for every skill. Finding the right combination of people matters most; a team with diverse, complementary personalities and talents makes all the difference.

A good data team should combine people with expertise across four main areas (“domains”). If I were starting a new data team, the first four people I would hire would be…

1. Business architect who knows your company’s business processes. Their success depends on convincing people to follow standard processes and solving data problems with minimal system changes.

2. Data architect who builds structures for data that make it useful for all future needs. They solve problems with their technical skills, grasp of database theory, and knowledge of best practices.

3. Application architect who understands how people interact with data and how they use software—like Tableau, MS Office, and AI/ML—to meet their needs. For them, success comes from solving problems creatively with user-facing tools.

4. Infrastructure architect who understands system administration, hosting alternatives, identity management, security, and how the unique workloads of analytic processing impact these areas. They’re measured by overall system performance, how it runs day to day. They’ll say “no” to solutions the system isn’t prepared to support.

Separating Roles

Intentionally draw clear lines defining each role. A business architect shouldn’t write any code, and an infrastructure architect shouldn’t create any objects in the database. Define clear quality control workflows for your project development. Your database architect, for example, should approve all development projects that include data pipeline code.

Here are some benefits of separating roles like this:

Checks and Balances.

Separating responsibilities eliminates conflicting goals. The “jack of all trades” approach that many companies use for their data team favors one person’s favorite solutions, with less focus on alternatives. As the old adage says, everything looks like a nail to someone who only owns a hammer.

Sustainability.

One of the most serious risks IT leaders face is a “single point of failure,” where all the technical knowledge for a solution or an entire system rests with one individual. This is especially common on data teams. When you separate the work like this, you’ve built shared knowledge and sustainability into every solution. Fewer people might be coding, but more people understand what that code does. Understanding the purpose of code matters a lot more than the syntax.

Creativity.

Data work requires vision; you’re always working toward something that doesn’t exist. Separating roles like this automatically brings more ideation to every project; people ask more questions about the problem (and its solution) because they come at it from different points of view. Everyone’s vision grows when more than one person controls the design.

Interdependence.

Separating roles like this creates what I call “interdependence”; people are forced to work together.

For example, they share technologies. You might notice that I haven’t said much about the specific software technologies your data team uses. In my management model, the whole team owns and uses all of the technology platforms. For example, every member of the data team knows how to use Tableau dashboards and SQL Server Integration Services; everyone has access to the development tools and understands their purpose within the platform architecture.

You avoid aligning tools with specific people and creating a single point of failure when things go wrong. When everyone uses a tool, the team can find more use cases for it. It makes tool selection a shared decision. Many data teams accumulate more and more software platforms over time, eventually owning almost every data management platform on the market.

You can always take the safe route and build a purely technical team. Nobody will question you. But that won’t set your team or company up for business success. Instead, use this “domain” approach to intentionally build a team with all the perspectives needed to truly make a difference across your company’s decision-making processes.

To remind you of this week’s data concept, enjoy Making Plans for Nigel, by XTC, from the Frictionless Data Spotify playlist.

For the full story about making data flow faster and better, check out Frictionless Data on Amazon.

Frictionless Decisions brings you counterintuitive, original, jargon-free ideas for connecting data to decisions. Every week.

Subscribe now

New technologies form us by the way they process and present information; they shape us into their image. We know this.

It took eighty years for the telegraph and photography to give us television, but then it changed every part of American life into entertainment, almost overnight. Forty years later, digital photos made us look better than we did in real life; now (twenty years later), AI makes us sound smarter than we really are.

“The takeover is complete,” says bestselling author Nicholas Carr.

I’m hearing this message across the board, from the very people working inside The Machine: thoughtful, intentional resistance will keep you sane - and human. You can’t avoid technology, but you can certainly resist the ways it forms you. AI shapes us faster and more deeply than its predecessors.

That’s the focus of this month’s Frictionless Decisions news roundup. The articles below suggest a deeper, more thoughtful approach than most. Digesting all this takes a lot of time (which you probably don’t have), so I’ve curated the best articles and posts I’ve read on this topic for you.

Give them a read!

Metannoying, by Hollis Robbins, Anecdotal Value, February 26, 2026.

Did you every wonder why AI writing is so boring? Robbins shows you patterns in AI writing and explains why they have this numbing effect on your mind. Maybe after reading this, AI writing will annoy you just as much as it bugs me.

If you don’t read this article, at least enjoy these brilliant notes from a great writer:

• “LLM-generated prose wants so much to be clear to everyone that it talks down to readers who don’t need the extra words, correcting beliefs they don’t hold, resolving contrasts they’ve already computed.”

• ”The ‘corrective contrast’ construction reduces ambiguity in the output space. Users want clarity. “Not X but Y” to the LLM is an insurance policy on clarity. Luckily for the designers of LLMs, corrective contrast also sounds cool, memorable, and often profound, at least in moderation. “I came not to bring peace but a sword.” (Matthew 10:34). Or “it’s not the heat it’s the humidity.”

Habits for Humanity in the Age of AI by Ben Sasse, The Wall Street Journal, May 8, 2026.

If you haven’t been following the former US Senator’s story, use this article to jump in. It’s happening in real time. Don’t miss his thoughts on technology.

How to Stay Human in an AI World by Pete Nicholas, The Gospel Coalition, May 13, 2026.

I’m not a fan of pastors commenting on AI, but this is a good take. Nicholas doesn’t pretend to explain AI, he simply focuses entirely on age-old habits that keep you human. It starts with fellowship…

The Bitter Lesson (of Decision Making), Data Tinkerer, April 30, 2026.

Machines love races; they always win. Let’s make that more specific: compute power always crushes the humans in the loop. And the power curve continues to accelerate. Maybe we should just get better at feeding The Machine, as the author encourages.

Is this a hopeless future? Well, yes - depending where you place your hope. Also read The Bitter Lesson for an even more bleak assessment.

What Leading a Data Team Actually Looks Like Right Now, Seattle Data Guy, May 14, 2026

I just love Ben’s street-level take on data management: “Data teams continue to be a catch-all team. Automations, AI, Dashboards, Analytics, Machine Learning, Data Apps. They can do it all, folks.” Or, at least that’s the expectation. Sounds exhausting, right? Keep reading!

Why Does AI Coding Make You Feel 10x More Productive And Exhausted, by Alejandro Aboy, Career Query Lessons, May 17, 2026.

It’s much better to listen to these data leaders than pundits like me! I love how leaders like Ale and Ben think; no pretention, only authentic advice and perspective. Yeah, automation promises freedom but delivers exhaustion. I was reminded of that yesterday when my addiction to Apple Maps got me lost for the 9th time this month.

Batter Up: AI, Labor, and the Policy Gap Nobody Is Closing by John Hughes, Modern Data 101, April 30, 2026.

What do you think about Human In The Loop (HITL)? I filed this article under my “manage it and it will be okay” category; the author uses the new Major League Baseball strike zone rules as an example of good “human-in-the-loop” practices. Is AI simply a threat to jobs? Is policy a solution? Or is it something darker?

The SaaS Apocalypse Fallacy: Why AI Raises the Value of the SaaS Bundle by Yakov Shkolnikov, Social Science Research Network, April 27, 2026.

CIOs, have you wondered about the hidden cost of using AI tools to build in-house applications? Yakov unearths these for you. Take this dose of realism to heart before your next application decision. “AI coding tools make the build path look dramatically cheaper and 2x faster than it is, which encourages organizations to skip the discipline a normal build-versus-buy analysis would impose,” he argues. It’s not just an opinion - Yakov proves it with academic research.

To remind you of this week’s data concept, enjoy The Voice, by Ultravox, from the Frictionless Data Spotify playlist.

For the full story about making data flow faster and better, check out Frictionless Data on Amazon.

Frictionless Decisions brings you counterintuitive, original, jargon-free ideas for connecting data to decisions. Every week.

Share

Of all the terrible ideas American carmakers ever had, their worst were in the 1970s.

Ford’s worst idea was, obviously, the Pinto, which earned a reputation for exploding after a simple rear-end collision. Even worse, Ford’s Mustang II wasn’t a Mustang at all; it was just a Pinto with a new logo. I could go on...listing cars like the Maverick and Zephyr. Ford couldn’t help themselves. Even their commercials were awful:

Other U.S. carmakers also forgot how to innovate in the 70s. General Motors’ 1971 “X-body” compact car stayed the same the whole decade but continually got new, cosmic-sounding names, like “The Buick Apollo”, “The Pontiac Ventura”, “The Phoenix”, “Oldsmobile Omega”, and “The Chevy Nova.” It finally ran out of gas as the 1979 Buick Skylark. The American Motors Company finished off the decade with the Gremlin (a shortened version of the Hornet).

The entire industry couldn’t get out of its funk. Historians call this the “Malaise Era” of American car building. American buyers started to prefer Japanese cars over American autos, and by 1980, Japan finally passed the U.S. to become the world's biggest car producer.

In his classic 1982 management book, Out of the Crisis, W.E. Deming gets so frustrated with American business managers from that decade that he repeatedly says they “rooked” (or cheated) consumers; they chased short-term profits and failed to understand how their businesses worked. And he would know. He’d helped Japan’s industry rebuild itself from scratch after World War II, without all the habitual management practices that plagued American car companies.

Unlike their American counterparts, Japan’s car companies were prepared to deal with global changes, like the 1970s oil crisis and the growing environmental concerns of American consumers.

Guaranteed Non-Innovation

American business leaders were blinded by decades of inertia, and they lazily repeated solutions from the previous forty years. Deming writes, “What is the competitive advantage of merely managing the way your competitors are managing?” Managers who don’t understand how their business works fall into this trap, relying on “best practices” to make safe and easy decisions.

Have you heard of these popular “best practices” in the data world?

• Moving data to Snowflake (or Databricks or anywhere in the cloud) will help your company make better decisions.

• You need more sophisticated software to analyze data.

• You should automate the work of analysts.

• Software will make you a better people manager.

Here’s a more subtle example: most companies purposely avoid using their core systems to connect products or services with financial data. Instead of configuring their transaction systems to do this, they build elaborate solutions using their data platform and reporting systems to connect them. Consulting firms even tell them to do this.

As Deming says, following “best practices” like this guarantees that you won’t innovate. You should avoid that at all costs.

If those assumptions sound familiar to you, instead, you should try to think outside the box, like this:

• Moving your data to another data storage platform doesn’t improve your decisions any more than parking your car in a new garage makes it drive faster.

• Excel is the best data analysis tool known to mankind.

• You need deep knowledge about how people in your company think to help them make better decisions with data.

• Never let AI evaluate people’s job performance.

Starting From Scratch

Like Deming says, the innovative Japanese carmakers weren’t tied down by the traditions of the American car companies. I had the same starting-from-scratch experience when I worked at Broadcom, where innovation was the norm.

We synchronized financial master data with product master data, creating far-reaching visibility into business decisions. Almost no chip company does this today because they assume the speed of corporate acquisitions and product data will make that relationship difficult to maintain. But we didn’t know any better back then. We connected these core data sets across every system and acquired 51 companies over their first decade.

Here’s some practical advice: Start with the master data, like I did. That’s a real strategic project nobody will ask you to do. It’s also free. Understand the purpose of the master data, the business processes that manage it, and the data strategies that make it more valuable.

Carve your own path and avoid best practices at all costs.

To remind you of this week’s data concept, enjoy Passenger by Siouxie and the Banshees, from the Frictionless Data Spotify playlist.

For the full story about making data flow faster and better, check out Frictionless Data on Amazon.

Frictionless Decisions brings you counterintuitive, original, jargon-free ideas for connecting data to decisions. Every week.

Share

The smallest datasets - often found on the desktop of an executive - offer more information about where the company is headed than the huge transactional datasets that come from your enterprise systems. This “crowdsourced” data is hard to find, hard to manage, and discouraged by “best practices.”

But this plan data gives visibility into the expectations of your company’s leadership. Capturing this data removes friction from many decisions.

I learned to look for “unseen” data from the unlikeliest of people, a Japanese executive who’d only known videotape manufacturing. Before I share the technical side of managing plan data, let me introduce you to Mr. Yutaka Mori.

Shipping Junk

In his best-selling book Chip Wars, Chris Miller explains how, in the 80s and 90s, U.S. semiconductor companies were essentially “shipping junk.” The worst US-produced chips had failure rates more than 10 times those of Japanese chip makers, something I experienced first-hand.

Manufacturing breakdowns at the small semiconductor company I worked for kept happening over and over, for a new reason every time; we had recurring non-recurring problems. The semiconductors we made were cutting-edge, but our ability to consistently produce working products with advanced technology was seriously lacking. As the Cost Accounting manager, I knew this because I tracked the numbers. The company threw away about 40% of the product produced at our manufacturing plant due to quality issues. Management set a goal to make a big improvement…and only scrap 30%.

TDK Corporation, the Japanese company famous for VHS tapes, knew this problem when they acquired my company. They must have felt their low-tech manufacturing success would translate to high-tech computer hardware.

They sent in Mr. Mori.

When the first results came in after he arrived, all the managers congratulated each other. We’d achieved our 30% scrap goal. But Mr. Mori was deeply troubled. He walked into my office and, in his broken English, asked if I could measure the total dollar value of everything the factory threw away.

“That’s easy,” I answered, “I’ll just set the scrap goal to zero. The metric will show all the product we’re wasting.”

What’s The Plan?

Your company generates huge volumes of transaction data from its business systems, like inventory, sales orders, and manufacturing activity. This operational data reflects the decisions made by your customers and vendors.

What data represents the decisions of your company’s leadership?

Plan data, obviously. If you’re starting a brand-new company, it’s the only data you have. Plan data reflects the decisions leaders make about where to invest their company’s time and money. Ask your business leaders where they keep their plan data; you’ll probably find it in a spreadsheet, on their desktop.

I know that capturing this data in your data warehouse doesn’t sound as exciting as your next AI project, so I’ll try to spice it up for you by explaining how plan data behaves from a technical perspective. As I walk you through this, think about how it contrasts with all your other data sources.

To manage plan data well, consider its four characteristics:

1. Source (where it originates). Plan data comes from business managers’ inputs, not from your business applications. You don’t need cutting-edge extraction, transformation, or loading tools for this. Simply create a structure for people in your company to submit their plan data.

2. Velocity (how fast it changes). Plan data changes intermittently. Like an outfielder in a baseball game, you might not get any action until the ninth inning, but when the ball comes to you, it will be the most critical play of the game. Similarly, plan data changes only now and then, when new news comes in. Sometimes it’s predictable – you know the date of the next corporate review – but other times the changes come out of the blue.

3. Retention (how much you need to save). Point-in-time “snapshots” matter most for plan data. When fans tune into a baseball game, they usually don’t care about everything that happened during the game. They just want to see the highlights. Similarly, analysts focus mostly on comparisons to previous plan versions. Usually, the snapshots synchronize with the business review calendar.

4. Volume (how fast it grows). Plan data grows in lumps. A single snapshot of plan data may not be large, but analysts want lots of snapshots for comparisons. They’re always answering questions about changes from the previous plan. So, whenever a new comparison point comes up, the entire dataset is copied (replicated). Plan data tends to pile up in chunks.

Don’t Lose Control

Mr. Mori thought more like an entrepreneur than a corporate citizen. When I updated the plan data as he requested, the metrics showed we were wasting millions of dollars in inventory each quarter. He told me to send this chart to every member of his staff every week. Sure enough, this scorecard helped him drive a dramatic improvement in factory yields.

Even though he knew nothing about the technical side of managing data, Mr. Mori understood that targets are the only data an executive truly controls. If you don’t capture executive plan data in your data platform, business teams manage it offline and no longer look to your system as the source of truth. You lose control of the data.

Don’t lose control of plan data. Get to know your executives, learn how they think, and manage their data.

To remind you of this week’s data concept, enjoy Making Plans for Nigel, by XTC, from the Frictionless Data Spotify playlist.

For the full story about making data flow faster and better, check out Frictionless Data on Amazon.

This article is part 2 of my series, “Write Slow, Think Deep,” Read Part 1 here.

Share

Everyone loses when writing becomes a race: attention spans shorten, ideas shrink, and everything becomes entertainment. Except the machines - they love races. They always win, and they want to take over all your writing. They’ll steal part of your soul in the process.

But slow, contemplative writing helps you internalize ideas and make them your own. That’s a lost art, as I wrote last week. You can’t avoid technology, but you can resist the ways it forms you, and I want to show you a practical way to do this. It won’t be easy.

The race is on, everywhere, including your own workplace. Consider just how widespread this epidemic has become:

The Bloggers

One influencer says, “If you’re posting 2 times a day, you should ask how to publish 3 times a day.” He offers a 7-step system for writing Substack articles five times faster (for $7). Here’s a quick summary of his summary:

1. First, brain dump your raw thoughts into AI. There’s “no need to think about structure, what comes first, what comes second,” or how the ideas you’ve dumped all connect together.

2. Are your raw thoughts incoherent? No problem, Claude solves all of this.

3. Inject some personality at the last step, and you’re ready to post.

Who reads all this high-speed content?

The Journalists

Nick Lichtenberg, a journalist for Fortune Magazine, used AI to generate 600 stories in nine months, twice as many articles as any of his colleagues. His process gives new meaning to the hot take: simple news stories, like one about Disney’s selection of a new CEO, take 10 minutes start-to-finish to get online.

Fortune Magazine is not large enough to compete with “publication of record” companies like the Wall Street Journal, but it must outperform (attract more traffic than) other massive content sources like Medium and Substack. Like other smaller publications, their survival depends on more writing volume, generated faster and faster. Overall, the number of AI-generated articles on the web surpassed human-written ones in late 2024, according to research and marketing agency Graphite.

Is it any surprise that our understanding of news grows shallower and more polarized?

Your HR department

It’s not just bloggers and journalists racing to write faster; even your own HR department wants to help you speed up your writing. They’ve got automation for your performance reviews, which I’m convinced most people will love. I took an informal survey before writing this article and couldn’t find anyone who doesn’t hate this writing assignment.

To almost everyone’s relief, Workday, the global workforce management software company, has built in an AI tool to write your employees’ performance reviews for you. Instead of spending time trying to remember everything an employee did, the application searches through all available project information and provides an employee assessment.

“AI frees managers from the burdens of manual documentation and subjective assessment, turning them into high-value coaches,” they promise.

That’s quite a claim. Can software really turn you into a high-value coach?

A Candidate for Slow Writing

If you think all this sounds completely exhausting, I’m right there with you. I don’t think you use AI for writing because you love it; you’re just in a hurry. But occasionally, writing really does become your main job at performance review time.

What makes an employee’s performance review different from those high-speed, high-volume examples?

You write a performance review for a person - a specific person you know. It’s written from your point of view, and nobody else’s. People’s livelihoods depend on your assessment. The message impacts your relationship with the person it’s written to. It may affect their future opportunities. You sign it in ink (hopefully) at the bottom.

Years ago - long before AI tools - Workday offered writing templates for performance evaluations. You’d select the tone (positive, slightly positive, neutral, etc.), and the system offered three or four sentences to use for each review question on the form. Maybe you’d personalize it, but there was a real chance that your own performance review would include the exact words your colleagues received on their review.

The templates solved a real problem; writing performance evaluations takes a long time. It’s hard, tedious, thinking work.

When you finally present the review to your employee, all they really want to know is the amount of their raise.

Internalizing Feedback

Managing people means helping them. It’s an obvious point, if you think about it, but it took me half my career to understand it. I was usually the last manager to complete performance reviews.

When I changed jobs mid-career, I finally realized that my technical skills weren’t the most important part of my role. A dozen people I’d never known now looked to me for their performance assessment. This reversed my thinking about the task; if I treated reviews as a compliance-driven formality, I would completely abdicate this responsibility.

I decided to invest a lot more time writing and forming the reviews. My goal was not only to give the most accurate assessment possible, but also to offer every staff member my best advice and a commitment to helping direct their future success. The writing process took a week or more, as I developed my understanding of each person’s talents and explored ideas and projects they could use to develop their careers. Since I’d internalized these ideas, I was able to recall them, discuss them, and build on them throughout the year. Feedback became a regular part of our conversations.

Deciding to do this came at a cost: spending more time mentoring my staff meant sacrificing time developing my own technical skills. I’d need to rely on others more.

If Workday is any indication, AI writes most people’s performance reviews now. A friend confirmed this recently when he told me he’d completed reviews for his entire staff in one morning.

“Nobody reads them anyway,” he told me.

I suggested he give the employees the AI prompt he’d used instead of the review itself.

Take The Long Way Home

Don’t let writing become a race. AI writing tools fill a page with lots of words, really fast. They take a good desire - expressing yourself, sharing your ideas - and give you a shortcut. In doing this, you miss the real value of the writing process: internalizing the ideas.

When you write manually, ideas work their way inside you and become your beliefs and values. When you write slowly, you think deeply. People recognize this and listen to your feedback.

It might even help them win the races that matter.

To remind you of this week’s data concept, enjoy Working in a Coal Mine, by Devo from the Frictionless Data Spotify playlist.

For the full story about making data flow faster and better, check out Frictionless Data on Amazon.

This article is part 2 of my 5-part series, “Data Lessons from Semiconductors.” Part 1 was Shrinking Transistors, Growing Insights.

Share

Talk about connecting the dots. Way back in the 1950s, Gordon Moore saw a connection between the tiniest electrical device - the transistor - and people’s willingness to hand over all their thinking to machines. Fiction writers in his day (like Bradbury) imagined this future, but Moore connected it to hardware.

When he wrote his famous “Moore’s Law” prediction about shrinking transistors in 1965, he used a cartoon to drive home his point. It showed a salesman pitching personal computers along with cosmetics and sewing products.

Did Moore really think this illustration would explain the greatest technological prediction of the century?1

Cartoon image from Gordon Moore’s 1965 article in Electronics Magazine.

Give It to the Machines

A semiconductor is the core piece of hardware that computers and digital products use to process data. These small “chips” of silicon pack as many electronic transistors as possible into a small plastic package that fits into your computer, phone, watch, or (fill in the blank).

Semiconductors were expensive to build and purchased mostly by the US military when Moore wrote his famous article for Electronics Magazine. A typical chip sold for about $30. If you translate that amount into the cost of computing power, it’s about a billion times more than an AI chip today.

The simple idea behind Moore’s prediction was that the cost per transistor - the tiny switch at the core of computer logic - was inversely proportional to the number of transistors on a chip. If a chip costs $1,000 to build, the cost per transistor would be $10 for a chip with 100 transistors. Double the transistor count to 200 on the chip (by shrinking them) and the cost per transistor drops to $5 each. That linear relationship between transistor count and cost has roughly held true for sixty years.

Moore’s point wasn’t about saving money. The way people lived would change forever, as computing power suddenly became less expensive; machines would take over most human activities that required logic. Gordon Moore anticipated our mad rush to create thinking machines, but even he would be shocked to see just how powerful and enduring that cultural mindset would become.

Wired Magazine brilliantly captured this story: “His breakthrough insight was that the pursuit of the best, lowest-cost electronics, motivated by economic competition, would necessarily create this doubling through an ongoing, extensive, and expensive social effort.”

Only a cartoon could capture the magnitude of this insight.

Pricing on the Learning Curve

Moore and the seven other electrical engineers who founded Fairchild Semiconductor had made huge progress in shrinking transistors since the era of vacuum-tube computers.

Bob Noyce, Fairchild’s CEO, didn’t think of Moore’s Law as an interesting scientific prediction; it demanded action. Selling chips to the military was the obvious way to make money - in the short term. It offered high margins and steady demand, but that business had a low ceiling. He made a big bet on Moore’s Law: he dropped the price of 709 Operational Amplifier chips to far below the cost of making them.

Almost nobody, aside from those Fairchild engineers, realized in 1965 how much this pricing decision would change the world.

Moore tells the story: “Bob got up and announced that Fairchild was going to sell integrated circuits for a dollar apiece. At that time, we were selling them for many times that, and they were costing us several dollars to make…There was an audible gasp in the room. People just couldn’t believe it. You have to remember, at that time, a single high-quality discrete transistor could cost a dollar or more. Here was Bob offering a complete circuit—with a dozen or more transistors and all the resistors—for the price of one component.”

“It was a bold move. It was pricing on the learning curve with a vengeance. Bob realized that the only way to get the volume we needed to make the technology cheap was to make it cheap first. He committed us to a path where we had to get the yields up or we’d lose our shirts.”

Moore often said that Noyce’s decision to price that chip at a dollar was as important as the invention of the integrated circuit itself. New markets and the transformation of human life wouldn’t happen without this courageous decision.

The Breakneck Race

The combination of shrinking transistors and the willingness of semiconductor leaders to bet on future innovation kicked off a torrid competition. Companies raced to find new applications for machines to take over human logic. It was a perfect combination of science and art, a race that continues to accelerate today.

Do you see the implication for data strategies? Business decisions would be forced to keep up with Moore’s Law.

Semiconductor companies with the ability to make fast, accurate, trusted business decisions at scale would win. Using data for those decisions wasn’t a nice-to-have option, and no semiconductor company understood this better than Broadcom in the late 1990s.

That’s where Frictionless Data was born.

To remind you of this week’s data concept, enjoy Wild, Wild Life by Talking Heads, from the Frictionless Data Spotify playlist.

For the full story about making data flow faster and better, check out Frictionless Data on Amazon.

Frictionless Decisions brings you counterintuitive, original, jargon-free ideas for connecting data to decisions. Every week.

Subscribe now

I want to tell you something, dear data leader, on behalf of all your users: we’re not sure you know us. We don’t all use data the same way. We don’t even ask the same kind of questions. We prefer different tools. You should try to understand how we think and work before building a system for us.

One-Size-Fits-All Data

This wasn’t obvious back in the day, when I started my career as an analyst.

At Texas Instruments, my first tech job, they called the accounting system “MODPLAN,” which I guessed stood for “modern plan,” but I’m not sure because there were no instructions or documentation. Everyone else – the people who’d worked at TI most of their lives – had figured out how to use it long ago and knew all its idiosyncrasies. The original system designer didn’t have somebody new like me in mind.

That got me into trouble.

My boss (the division president) convinced another division to reimburse us for some work. Now, entering a “credit” journal entry in MODPLAN required putting a minus sign in front of the number, and not knowing this led me to book a $2 million entry in reverse. Instead of receiving the credit, I gave the money away. The only reason I didn’t lose my job was that they had already laid off everyone else.

Business software developers in the 1990s often let a key stakeholder create a solution tailored mostly to their personal tastes. They fell into this common trap, leading to some very strange solutions. Ironically, the software tools offered for data analysis today suffer from a similar problem: one-size-fits-all solutions.

Creating More Work

They told you their project priorities, you took notes, and now you think you know your customers. You know what they want, but you don’t understand how they think.

Here’s a test: try following the data trail all the way from your data sources to an executive’s desktop. Typically, data flows like this:

You think you’re helping the analysts, but you’re really giving them work to do, and your one-size-fits-all solutions just prove to them that you don’t understand.

It’s like going to cap night at the local baseball stadium. I love baseball, but I wear a larger size than most people, and those free baseball hats make me look like a cone head. One-size-fits-all solutions always leave someone looking foolish.

Data Personas

In the 1990s, tech pioneer Alan Cooper proposed a new approach to software development: the “user experience movement.” He defined software requirements for fictional characters who represented the typical person in a job function, rather than the software developer’s closest friend. These template users, which Cooper called “personas,” helped standardize business processes and scale up software usage.

Three types of people use data for business decisions: Executives, Analysts, and Service Staff.

People in these roles ask different types of questions from different points of view at different frequencies. Some ask questions about business transactions; others ask about the future of the company as a whole. Some ask questions about the totals; others go directly to the details. Some ask the same question every day; others don’t know their question until they see the data.

Here’s a quick summary of three types of people who use data for decisions:

Table 1: Data interactions by persona

Executives

Most people identify an “executive” by their job title – a Vice President, General Manager, Senior Director, Chief Financial Officer, or Chief Anything. They spend most of their time managing the work rather than executing it. They track a predetermined set of metrics, viewing data from a top-down, aggregate perspective. They ask questions like, “Are we measuring the right thing?” or “Is today’s total what I expected?”

They ask similar questions day after day; they know where they want the company to go and want to quickly verify whether things are on track.

Service Staff

Think of a person who interacts with customers and handles the process and issues related to sales order processing, or a purchasing staff member who helps employees find and buy the supplies they need to do the work. These people care about transactions. They ask, “Do I have the data that I need to execute my job, when I need it?”, “Is the latest part version available?”, or “Is the customer zip code correct?”

Operational people use data records one at a time, moving a single business transaction to the next step in the process. Data quality matters most to them because real actions, like shipping a product or locating an item in the warehouse, depend on it. Everyone depends on them, and they depend on you.

Analysts

Analysts bridge the gap between executive questions and what’s happening with the service staff.

They gather data, interpret it, draw conclusions, and recommend actions. They work in all parts of the business – sales, finance, operations, or marketing - with job titles like business analyst, financial analyst, (fill in the blank) analyst, or even manager. They’re usually hands-on with data tools, such as online queries, dashboards, downloads, and spreadsheets.

And they’ve got to be ready for anything. They don’t know what data they need until someone asks them a question. When it comes, they investigate and explore the data. What they find forms the story they’ll read back to the exec, and they don’t know how far the line of questioning will go.

Get to know your customers by learning how they think. They’re not all the same. Make sure everyone gets a hat, and everyone’s hat fits them.

To remind you of this week’s data concept, enjoy Save It For Laterby The English Beat, from the Frictionless Data Spotify playlist.

For the full story about making data flow faster and better, check out Frictionless Data on Amazon.