AI answers pull heavily from community content. The platforms and the amount of content used are fluid, specific to engines/models, and dependent on agreements (or disputes) that broker access.

What isn’t in dispute is that community-generated content is, and will remain, an important source for a wide range of AI.

The study, intended for experienced community and CX practitioners, explores:

• Which third-party platforms and communities matter to how their brand is represented in AI answers;

• Which of those they actually participate in;

• What success looks like to leadership, and what metrics are used to track this;

• What sources are being cited;

• Which tools teams use to measure, monitor, and support workflows;

• Who owns the work internally, and how it's resourced;

• Which governance capabilities are in place and which proved useful.

Our intent is to establish a baseline for a practice that doesn't have one: what community teams are doing about AI-generated answers, who is responsible for them, and whether they connect to how they already make the case for community.

Take the survey here:
https://structure3c.typeform.com/AEO-Community26

Respondents will receive a summary copy of the results, which we aim to release in mid-October.

Please note: we reserve the right to (and likely will) filter platform, agency, and consultant answers from the results and summary.

I first saw the internet on television.

It was my senior year of college, 1995, and PBS ran a segment about something called a web browser. Mosaic. I was floored. I went to the library the next day, got myself a VAX account, and started exploring.

I was an art major, working on a BFA in graphic design, which meant I had spent four years in the middle of a transition from mechanical photo reproduction to digital. Everything I was learning about the practice of design and how images were made and reproduced was being replaced while I was learning it. I had already experienced the ground moving beneath my feet. The web was a whole different level.

What fascinated me, and what has rattled around in my head ever since, was the underlying premise: the internet connects any two things on the network - people, content, machines - and the distance between them is effectively zero. Not metaphorically zero. Structurally zero. You could hypothetically reach anyone or anything, and the network didn’t care about the hierarchy or geography or institution in between.

That seemed like the most important idea I had ever encountered.

Four years later, I had an Internet job.

I joined TechRepublic in 1999, when it was still called Narrowcast Technologies. The premise was a hyper-personalized content portal and community for IT professionals, everyone from the support desk up to the CIO. The executives were mostly ex-Microsoft people, experienced with content newsletters and with TechNet. They understood that IT pros were an audience (and potential community) that wanted to talk to each other, not be marketed to. What was new was that now you could actually let them.

At some point in early 2000, I was handed a copy of the Cluetrain Manifesto.

If you’ve never read it: four guys nailed ninety-five theses to the web in March of 1999, arguing that markets are conversations, that hyperlinks subvert hierarchy, and that companies had better start talking like human beings because their customers already were.

The buzzword was disintermediation. Cut out the middle. Make the direct connection as direct as possible. Let people inside companies talk to people outside companies without it being routed through PR or Communications first. It was the zero-distance idea applied to business, and it landed in an industry full of people who wanted to believe it.

It was really a heady time. The customer was going to be in charge. The internet was going to be the catalyst that shifted power back from institutions and hierarchies to people. It genuinely felt like anything was possible, and the collective we (the whole industry) pushed hard on that, in every direction at once. Amazon. The early social networks that eventually inspired Facebook and LinkedIn. Hosted communities like the one I was working on.

Here’s the part I think about now: The reason Cluetrain worked wasn’t that it was right. Plenty of things were right… and plenty of things in hindsight are wrong. The writing style is a little cringe. But it worked because it was aimed at the generation of tech founders and executives who were looking for a thesis about how network technology would disrupt business, and who had budget.

That turns out to matter. Transformation ideas don’t get traction because they’re true. They become true when someone with a budget adopts them.

Looking back, I’ve tried to arrange evolutionary periods I’ve worked with into eras. I don’t find hindsight to actually be 20/20, and some of this is recent enough that I’m still working it out. The early definitions feel solid to me. The later ones are more malleable.

The early web, roughly 1995 to 2002. Very few people were online, and whatever door you came in through was often your entire experience of the thing. America Online. The WELL. These were containers for the internet more than windows onto it, and they were organized as communities, people, and discussions at the same level of priority as content.

The social networking boom. MySpace. LinkedIn. Blogging, which was just people writing journals in public, scaled up to challenge mainstream media. It turned out to be the most consequential of the three. Then Facebook, Twitter, YouTube.

Mobile social. The iPhone launched in 2007, and the Facebook iOS app launched in July of 2008. We had a monolithic, magical device in our pockets, wonderful to handle and beautiful to look at, that we kept on us twenty-four hours a day and that knew where we were on a map in relation to everyone else. It was inevitable that the social networks would manifest there and take advantage of all of it.

I want to stay on that transition for a minute, because I was standing in a specific place while it happened.

I was at Forum One then, running conferences in the Bay Area. In 2007, the people who came to those conferences were executives and leaders whose business models had community as a core principle, in the traditional sense of the word: bring people together so they can create value with each other, give them the ability and space to actually get to know one another, and give them a lot of agency over what they do, what they see, and what the community experience is like.

Over the next few years, that audience was complemented by, and then overtaken by, people who worked at social networks.

Because there was a gold rush on. From roughly 2008 through 2012, an enormous amount of money went into building new platforms whose actual business model was advertising, or selling user data, or both. That was the revenue strategy. And somewhere in there the more benevolent idea of community got thrown out the window.

Zuckerberg talked about Facebook as a community in the early days. I don’t know whether he was naive or disingenuous, and I don’t especially want to relitigate it. I do know that the word survived and the practice was harmed.

Social media sucked all the oxygen, attention, and budget out of the rooms community builders were in. It did that for about a decade.

Which is why Dell was strange.

When I went there in spring of 2010, it was one of the very few large companies genuinely trying to build a community ecosystem on the open web. Was there a profit motive? Eventually, yes, obviously. But it started as experimentation, and as a real attempt to understand how a company that size could create actual relationships with customers, create value for them, and increase its own empathy… and of course, profits, in the process.

I came in as Director of Global Community and a founding member of the Social Media and Community team, which everyone called “SMaC”. There were thirteen of us at director level, each owning a set of business functions across community and social. We reported to a VP, who reported to a board inside Dell that included the executive team. Including Michael.

The VP pulled all of us to Austin for a two-day offsite. In two days, in one room, with remarkably little friction, we built the business plan and the program set for the next two years of social and community strategy at Dell.

He also had a mantra, which I’ve stolen and used ever since. Companies breathe in and breathe out when they need to make change. Dell was breathing in: pulling social and community decision-making and strategy into one place, one team, one plan. And we had two years before the company exhaled and distributed us back out into the business.

That was the design, and the functional dispersal was the point. Two years to build the strategy, the policy, the governance, the programs, and the partnerships, and then thirteen directors get placed into senior roles across the business units and carry it with them. I went to Large Enterprise as Director of Social Media and Community.

This is worth repeating: the plan was to build a center of gravity and then deliberately push the pieces out into the larger org.

The organizing idea was ecosystem.

We knew customers weren’t in one place. They were spread across the web based on what they needed, what they were interested in, and their preferences: some people want a forum, some people want to complain on Twitter, some people want a private channel with a rep and nothing else. We didn’t stop to argue about which one was the best. The ecosystem was the lens, and it played out across different form factors: social media, hosted community, and more private collaborative spaces.

Four things sat inside that.

Hosted community. The best community experience we could build, plus a real understanding of where the value was actually coming from: support deflection, customer knowledge, advocacy, and innovation through IdeaStorm, which was our ideation portal. I also started the Dell Rockstars program, which was our advocacy program for the customers who were making extraordinary community contributions.

Social media at scale. A large team covering global social media, and beyond that, the strategy and the resources so that the rest of the company could actually show up and participate.

Listening. We pioneered a lot of the technology and a lot of the ideas for monitoring the social web at scale, globally. Including China, which was emerging then and which was fundamentally different from the Western web in ways that have since played out in public.

Social innovation. The most interesting thing we built here was essentially an on-demand Net Promoter Score, running off the listening data. Real-time social sentiment, which let us do a number of things, including reaching out directly to accounts that needed it before they asked.

We also spent an enormous amount of time instrumenting things.

All of our own digital properties, and as many of the social properties as we could reach. This was an early attempt at what would now be called a universal customer profile. The goal was to know, at minimum, which account a person was coming from, even if we couldn’t identify the individual.

That tooling was advanced for the time, and it let us answer questions nobody in a large company had been able to answer before. We could track when social media or community appeared in the purchase path and report that number as social or community-influenced revenue. We could correlate how often someone participated in the community, or used community content, against purchase size and frequency, and get at the net effect of community participation on buying decisions.

The one I’m proudest of was the IdeaStorm study. We identified five ideas that the product teams agreed had been solely sourced and formed on IdeaStorm and then brought to market. We could have found more. We stopped at five because five was enough to make the point. Then we studied the actual revenue from those products and compared it against what we’d spent on IdeaStorm.

That got us a real number, and it got me a “get stuck in the elevator with an Exec” stat, which is what you actually need in a large company. Grossly dividing the revenue across all the ideas ever submitted, every idea on IdeaStorm was worth approximately ten thousand dollars.

Why am I going into the details? Because I want you to understand the breadth and depth of what a Fortune 50 company was willing to fund in 2011, and the level of proof we were expected to produce.

It may sound non-intuitive, but the community of supporters you build inside the organization is just as important as the community of customers and partners you are building outside the organization. We did two key things to establish and strengthen stakeholder relationships.

The first was the partnership structure. Every one of the thirteen directors, me included, had a counterpart in the business who we met with regularly. The job was to course-correct their social and community strategy, understand what products were coming, and figure out how we could support them better. We were essentially very senior account managers to our own company. It built genuinely strong relationships with colleagues in the business units, and it’s the single piece of the model I’ve tried hardest to replicate with clients since.

The second was communications, which was more sophisticated than anything I’d seen. At the top end, short newsletters to VPs and directors — the people who actually held budget in the lines of business — about impact. During the peak, those went out weekly. Demonstrate impact, tell the story, keep the support. Once the listening tools were in place, we also set up dashboards for specific executives so they had a live view of what was being said about Dell, roughly how people felt about it, and what was coming of it.

And then there was the part I still think was the most interesting thing we did.

Once the strategy was set, we trained and equipped Dell employees to participate on behalf of the brand. There were levels of SMaC certification [see below], and the point was to enable the people who wanted to go out into the ecosystem and have the conversations they were already interested in having, in the places they already were.

What we were really doing was walking willing employees toward the edge of the corporate membrane. Out to the boundary, where the company touches everyone and everything else.

Which is a lot to ask of someone. So we built a fairly robust escalation path underneath it for support and communications issues, so no single employee ever felt hung out to dry or ganged up on out there by themselves.

Watching the team behind the SMaC training run the program all over the world was impressive. I was proud of the part I got to play to help build our internal community of practitioners.

I’d done a lot of unconferences before Dell, so I suggested the idea of Unconferences after some of our larger SMaC training sessions to allow employees to discuss what they had learned and explore ways of putting the strategies and tactics into action. The SMaC unconference was born! The first one was in Round Rock, Texas, with a couple hundred people. Michael showed up at the end to talk about the importance of being close to our customers and to endorse the program.

We ran them all over the world after that. The one in Xiamen was supported by Chinese & English translators working live, and we held the whole conversation in real time in two languages.

None of this happened in the dark.

We briefed Forrester and Gartner constantly. Analysts came through, wrote us up, used us as the example. People writing books about social business studied what we were doing. I never wrote one of those books. I was in the room the books were written about.

Forrester ran a program called the Groundswell Awards to recognize transformational social business programs. Dell won at the first one in 2007, three years before I arrived (which is part of what attracted me in the first place), and won again in 2011 for the Listening Command Center.

Not long after, I went back to Autodesk, bringing all the Dell learning and experiences with me. In 2015, they won a Groundswell Award for Expert Elite, an advocacy program modeled on, and extending the best “MVP” programs at the time, including the Dell Rockstars. It’s still running. It has now survived every era in this piece, and it is still producing value.

As far as I can tell, 2015 was also the last year of the Groundswell Awards.

In 2012, Dell exhaled.

The thirteen of us went out to our business units. A new round of Executives were cycling in to run Marketing and were more skeptical and less enthusiastic about the social investments and uncomfortable with the lack of control. The functions lived on for another two or three years, as far as I know. By then I was at Autodesk, and I genuinely don’t know what happened to all the pieces after that. Not a glamorous ending to this thread, but true.

As I was winding down my time at Autodesk and preparing to launch Structure, I started noticing something. The broad, generally accepted principle among executives that community and social were important investments (the thing that had made the previous five years possible) started to fade. This wasn’t a heated argument or a heavily debated topic, but a slow fade. Less oxygen in the room every year. We were in a down cycle of social media investment, and it took community investment with it.

Then the pandemic.

I’m not going to armchair quarterback the socio-political implications of the pandemic here, because it wouldn’t be fun or rewarding for either of us. But looking for the silver lining from a very dark cloud: by March of 2020, everyone who needed to interact with another human outside their immediate household was doing it on something like Zoom. Near-universal, near-simultaneous adoption of the collaboration technologies. And I think that reminded a lot of people of something they’d stopped prioritizing: that digital relationships have to be built and maintained, and that there is craft in doing it well.

Since 2020, we’ve been on a slow climb back. Part of that has been a reckoning with the good and the (largely) bad parts of mass social media. Part of it has been companies taking a second look at community investments they abandoned in the previous decade. The work I enjoy most now is helping clients figure out where they need to invest: what to host on their own domain, where to participate, where to show up in person, and where to show up online. That’s the community strategy, in a microcosm.

Here’s what I keep running into, though.

Some of the institutional knowledge still exists, in pockets. There are people who know how to do this. But most of the social media and community staff in enterprise has turned over by now. Some moved into new roles, and plenty left the practice entirely.

What’s frayed is more than infrastructure. It’s the coherent vision, the governance function, the programs, and yes, most definitely the infrastructure.

I have a client right now who had a robust social listening operation ten or twelve years ago. Instrumented, staffed, routed, the whole thing. I checked in with them last week. They are hand-building and vibe-coding a tool to track brand mentions on Reddit.

Not because they don’t know better. Because the thing that existed is gone and nobody who’s there now was there then.

There are lots of reasons. Executive churn is the one I’d put on top. The CMO usually holds a big chunk of this budget, and average CMO tenure is now 4.1 years, the shortest of any C-suite role except COO. So a program gets handed down through two or three executives, none of whom built it, each of whom needs a story of their own.

Underneath the churn there's a fluency problem. I have watched more than one company hand a social and community portfolio to an executive who had never used the platforms that mattered most to that business. That isn't a knock on them. They were hired for other things, and they were good at those things. But it tells you how the practice was valued. Fluency in it was never a qualification for owning it, which means every handoff is a coin flip on whether the next person learns it or defers to their staff.

And then sometimes the programs don’t get killed at all. They get outranked. Dell went private in 2013, the largest corporate privatization in history at the time, and then merged with EMC two years later. That is a decade of executive attention spent on existential-scale corporate events. I don’t think anyone in that building made a decision to abandon social business. I think it stopped being the most important thing in the room, and then it stopped being in the room.

Is it unusual for a program to get deprecated somewhere in there? No. It’s completely ordinary. I just think it’s unfortunate.

That slow climb since 2020 has mostly been belief. People remembered that customer relationships are valuable, and some budget followed. It’s real, but it’s soft, and soft things get cut.

What’s happened in the last year is a business case.

AI is commanding executive attention. Models need data to train on. So: guess where there’s a large supply of high-quality, human-written, topically organized, natively threaded text about how products actually work in practice.

Hosted communities. And community-shaped networks like Reddit.

The second piece is downstream of the first. Community and social content is showing up in answer engines, which means it’s showing up at the start of the buying process. Not the top of the funnel exactly (it isn’t really a funnel anymore), but at the moment someone forms a first impression of whether your product does the thing they need. That impression is now frequently assembled by a machine, out of text your customers wrote, in places you may or may not participate in.

What that adds up to is that companies with a real community strategy (programs, infrastructure, staff, and a clear-eyed view of what they’re actually trying to do) now have a measurable advantage over companies that don’t. Not a loyalty advantage, which was always the argument and was always slow to prove. A discovery advantage.

In enterprise, AI is where the money is actually moving. Every company I work with is spending on AI tooling and then discovering they need strategy and training to go with it. That turns out to be an unlock for community work, because the case for community now sits inside the case for AI instead of competing with it for the same dollar.

Which is a strange thing to sit with if you were around for previous cycles. We spent years arguing that companies should invest in community because customers deserve to be treated like people. That argument came back after 2020, and it’s still perceived as “soft”. The argument that’s working now is that community produces the training data and the citations. Same investment, mostly the same programs, an entirely colder reason.

The uncomfortable part is that it treats people as a supply of text. But the extractive version of this doesn't actually work over time. Community members and mods are pushing back on harvesting efforts (ex: Stack Overflow), and the danger is that good contributors leave and take the quality with them. If what you want is a corpus worth citing, the only way to get one is the slow way: real experts, real answers, real reciprocity. The expected outcomes from the “cold” argument turn out to require the investment of the “warm” argument: treating customers like humans.

So where are we standing, and what’s actually in front of us?

A little inside baseball first. The community platform market has been through a series of disruptions over the last two years after a long stretch of stagnant product vision and development, and most of them look positive to me. The major players are working through what AI actually does inside a community experience, and building toward what community can do beyond support.

Here’s what I see coming.

Hosted communities seen in a new light. The old argument was that communities drive loyalty and preference, and that hasn’t stopped being true. But the more compelling argument now is that the community you host is a corpus you influence and data you own. Given how people are actually finding products and forming opinions, that matters in a way it didn’t three years ago.

Local, and in person. We can’t think about community as an online-only thing anymore. What I’m seeing is renewed interest and renewed investment in local social networks and in real-world events, even when they’re coordinated through a national or global hub. Local has never been more important. And in-person as a form factor keeps gaining as screen fatigue grows, which is the one experience an agent can’t attend for you.

Communities doing work they’ve only been adjacent to. Community has spent twenty years proving itself in support and advocacy, which is where the measurement was easiest. The functions I’m interested in now sit further up the strategy stack: innovation, and strategic forecasting and futures work (we’re building this with a client right now), new forms of community-based product trials and structured learning, private industry networks offered as a value-add rather than a marketing channel. These aren’t exactly new ideas. IdeaStorm was an innovation program in 2007. What’s different is that the tooling can finally support them, and that a community capable of doing forecasting work with you is a very different asset than one that answers install questions.

Agents in communities. There are early signs of agents and personifications of AI showing up as participants. The extreme version is Moltbook, which launched in January as a social network exclusively for AI agents, where humans are permitted to observe but not post. It hit well over a million registered agents in about a week, and Meta bought it six weeks after launch. Another example is Buzz: agents holding their own identities and memberships inside a human workspace. The thing to watch is digital personas becoming first-ish class citizens in a network, and what that does to the norms of a place built for people.

Confederations. I think we’ll see brands partnering on shared community experiences to serve a whole customer rather than a slice of one. Imagine a financial services firm, a logistics company, and an office supply business jointly supporting a small business community in a specific vertical, instead of each of them separately trying to generate interest in their own owned community. That sits in the space between a mass platform owned by a third party and a fully hosted community. I don’t know of a live example at scale yet. I think it’s coming, and I think it’s the most interesting structural idea on this list.

Rebuilding the connective tissue. The thing that made Dell SMaC work at global scale wasn’t the programs or the technology. It was thirteen directors with named counterparts across every business unit, a governance function, and a communication rhythm that kept the company aware of what was happening. That’s what produces programs everyone benefits from instead of silos and fiefdoms. It’s also the first thing to go when a team gets cut, and the hardest thing to rebuild, because it isn’t a purchase order. You can buy listening infrastructure back in a quarter. You can’t buy back the cross-functional relationships, the competencies, or the trust.

Nobody has a crystal ball here.

We’re living in a standing wave of disruption. New models and new functionality appear weekly, sometimes daily, cascading through the products we use, the experiences we have, what shows up online as artifacts, and eventually what happens in the real world. Anyone claiming a coherent and comprehensive view of what’s actually happening globally is, objectively, full of it.

I’ll leave you with the thing I started with instead.

The original promise of the Internet was zero distance between any two people or things. That was a wonderful, inspiring vision. Turns out it was not a realistic one, and we didn’t think through the implications or the consequences in any meaningful way.

Thirty years later, it’s worth asking what is actually at zero distance, what is actually worth connecting with, and what is mediating everything else.

That’s the place we’re standing. Right here, right now. Do we build toward the original intent, less mediation and more direct connections between people and organizations that need each other? Or do we build a better and more pleasant version of the mediated enclosure?

The business case for community is being made by the AI budget. But it requires the same behavior we were arguing for in 2011: be porous, let your experts out to the edge of the corporate membrane, participate in places you don’t own, and give people reasons to keep contributing. Social business was mostly right about what to do. It was wrong about what would keep it funded. And now companies are having to relearn and rebuild.

So start with a community vision and strategy, not tools. Most companies can’t start with the people who remember, because those people are gone. What you can do is design what you’re actually trying to build before you buy anything, and then find enough leaders who want the same thing to bring it to life.

I keep coming back to the quote from Howard Rheingold about the WELL, one of the first online communities:

What it is, is up to us.

Ask a room full of executives why the business should invest in the community, and if you aren’t escorted out, you will get at least as many answers as there are people.

That is not necessarily a failure of the room, but rather a reflection of the subject's complex nature. Community investment is a full-contact sport: many stakeholders, competing priorities, real budget, and no shared vision, at least not one that lasts more than 3-4 years. Everyone wants to jump to what (the vision), where (the ecosystem), and how (the programs). Almost nobody stops to settle why. So the why never gets settled, and every initiative that follows winds up relitigating it. The “why” question keeps coming back because it is never actually answered.

So let’s try to answer it.

The familiar case is: Customers solve problems, learn from peers, find working examples, and get more out of what they already bought. Partners learn faster and see where they fit. Fine. But look at how most companies build on that truth. A dozen small programs chasing small groups for narrow outcomes. Three teams all touching the same customer or the same partner, none of them holding the whole relationship. Efficient? Strategic? It is neither. It is a bag of tactics masquerading as a strategy.

The usual critique is that companies get stuck measuring engagement, but that’s not quite right. Plenty of smart people already know measurements that would be classified as “engagement” only tell part of the story. The real problem is they do not know what to measure instead. The methods and the math for it are still emerging, so teams count what is easy to count. But the thing worth counting is whether the relationships around the business are becoming an asset with real, accruing value, over time.

Now the “why” question is urgent. AI is turbo-charging the fast layers of the business: content, commerce, transactions, the surface churn. Stewart Brand’s pace layers are a helpful tool here, especially the idea (simplified for our purposes) that the fast parts learn, the slow parts remember. Community is a slow layer, where trust and memory accumulate over years. Customers used to start at your website, or at least somewhere in your digital ecosystem. Now they find out about you in places you don’t own or even necessarily influence.

The reflex is to throw more content at the problem. The problem is that everyone now has access to an infinite content machine that can spray as much competent but forgettable content anywhere they wish. Information is no longer a scarce resource. Treating your customers and partners as raw material to feed the machine never ends well. We are all living through a shared social media hangover, and people are genuinely pushing back on being used by a tool instead of being in the role of tool user.

A key distinction: you cannot separate knowledge from the knower. Information sits in a database. Knowledge lives in a person, with all their scars, lived experiences, and idiosyncracies. A model can retrieve the information. What it cannot hand you is the tacit knowledge (Polanyi’s phrase, “we know more than we can tell”) of someone who already hit the exact wall you are about to hit, at a company like yours, under the same pressure.

This idea first clicked for me shortly after we launched TechRepublic.com in 1999. We had one of our first posts in a Windows server forum, which was an incredibly complex permissions issue. Within 12 hours, 2 other members from different parts of the world offered up parts of a solution that helped the original poster solve the problem. It occurred to me that we had helped bring together the few people in the world who had encountered this particular problem and could solve it.

I saw this happen at global scale with our extended team members at Dell in late 2010. Part of our SMaC (Social Media and Community) team’s mission was to train and enable any customer-facing employee to interact on our communities and social media ecosystem. Globally. As part of the training process, we hosted global Unconferences with no predefined agenda that allowed thousands of Dell employees to come together to share their real experiences, issues, and advice. Beyond static training modules, these Unconferences created a community of practice that lived on long beyond the events.

Here is the thing I see missing from most strategy discussions: the best communities advance people. Professionally, in their craft and their careers, and personally, in who they are becoming. That is the part no library and no model reproduces, because it is not information moving around. It is people developing. Wendell Berry’s word for this concept is membership: not a list you join, but a web of people who show up for each other.

So the asset was never just a platform, program, or group. It is the network of relationships the organization has made possible. Everything the CFO likes (retention, expansion, advocacy, product signal) is an outcome of those relationships being made stronger.

That changes what you measure. The trap is to reach for whatever is easy to count, which is almost always a transaction: tickets deflected, leads sourced, seats renewed. This approach measures the exhaust, not the engine performance. Instead, explore: are people developing, moving from the edge inward, from newcomer to contributor to the one others turn to? Is value moving sideways between members, or does it all still route through you? That is what an appreciating relationship looks like from the outside, over a longer time horizon than this quarter’s dashboard. The methods for measuring it well are still maturing, but IMHO, the direction is not in doubt.

I see two future roads from here. One treats community as a content supply chain for the machine, for search and marketing automation. The other treats it as a relationship system (layer/infrastructure/substrate) that helps people make better decisions and build things together. Relationships are the rare asset that appreciates and is somewhat portable. When your champion changes jobs, the relationship goes with her. The person and related relationships are the asset. The sold seat of software is the transaction.

That is the why. Or at least, my current thinking on it.

Everything else is what, where, and how.

Early in my career, a few months after we launched TechRepublic, I watched something happen in the community that helped me understand the possibilities that communities represent. Someone posted a very specific technical question about a problem most people would never encounter. Two members from outside the US, who were probably the only two people in the world equipped to answer it, did.

I was blown away. That was the power of our community, creating the conditions for collisions that help everyone advance together.

I thought about that moment again today on a panel at the All Things in Moderation conference. Venessa Paech asked which inevitability narrative about the future of community we should stop accepting. Mine was easy: the story that AI will automate everything.

We learned why that’s a fallacy the hard way when I was leading a community for a design software company. We were early in experimenting with agents. The models are good at one thing, ferreting out an answer that already exists somewhere. When a member hits a genuinely new problem out in the wild, something the model has never seen, it has nothing to offer. The collision I watched at TechRepublic was exactly that kind of problem. So the idea that AI will replace this work, or even keep a community’s knowledge base alive on its own, runs into a wall that has nothing to do with optimism or pessimism about the technology. It simply isn’t technically possible.

What a community actually produces is relationships, and those relationships outlast the people who tend them. They outlast the staff and the executives who fund the work, and they keep compounding value long after both have moved on. A model can summarize what a community has already said. It can’t carry a network of relationships for decades.

Which makes the current arrangement worth looking at, honestly. The most valuable members, the ones answering the impossible questions, are feeding what they know into systems that increasingly reuse it, and most of them never see the value they create or get much back for it. That exchange was already lopsided. As contributions get absorbed into models and resold, it tilts further. Any serious version of the future has to make the exchange visible and return something for it, rather than taking it opaquely.

This is what bothers me about how many tech leaders tell their story of a future that is fixed. It tells leaders the outcome is already decided, that the only question left is how fast to adopt. The future isn’t set. Howard Rheingold, writing about The WELL, one of the first online communities, captured something I keep coming back to, that “What it is, is up to us”, “it” being the practice of community building in this context. There are many futures ahead, and we get to (we must) steer toward the one we want.

The best version of AI in this work absorbs the drudgery so people can do the part that only people can do. The worst version is the one where we become the tool the tool uses, which is roughly where mass social media left us. What separates those two outcomes has little to do with the technology and everything to do with whether the people making strategy and design decisions about community today keep a hand on the tiller.

If I had to put one principle at the top of a charter for the next decade of this work, it would borrow from the Hippocratic oath: human well-being comes first, including the well-being of the people doing the work. Build from there, and build around possibilities rather than constraints.

To paraphrase Howard Rheingold, the future of community practice, participation, and platforms is up to us.

I am actively trying to understand why I still use mass social media.

I'm beyond the “is social media good or bad” debate. It’s exhausting. We have plenty of data showing, especially on most of the Meta platforms, that regular and extended use is bad.

A more productive way to think about this is to consider what we are actually trying to accomplish on any given platform. One helpful lens for this question is Jobs To Be Done.

Jobs To Be Done is a framework first introduced by Clayton Christensen in response to the habit of companies relying solely on market research data to understand their customers.

“After decades of watching great companies fail, we’ve come to the conclusion that the focus on correlation—and on knowing more and more about customers—is taking firms in the wrong direction. What they really need to home in on is the progress that the customer is trying to make in a given circumstance—what the customer hopes to accomplish. This is what we’ve come to call the job to be done.”

Know Your Customers’ “Jobs to Be Done” - Christensen, Hall, Dillon and Duncan - HBR Sept 2016

So what job are we “hiring” (a.k.a. choosing) any of the various mass social media platforms for? And the more important question: What have they implicitly hired us for? Let’s take a look.

LinkedIn. Strip away the feed and what's left is what we actually hire: a directory of working professionals, a resume of record, and an inbox where recruiters and prospects can find us. Nobody has ever had a struggling moment that broetry solved. The feed isn't the product; it's the tax we pay to use the directory.

Facebook. The original job, staying connected to the people we care about, was fired years ago and rehired by the group chat. What survived is pure utility: Marketplace, Groups, Events, and the birthdays we’d otherwise forget. Facebook didn’t die, it got demoted to the junk drawer of the internet, full of things we rarely think about but aren’t ready to throw out.

Reddit. We hire Reddit for one job: honest human judgment from people with no commercial stake in the answer. The proof is the word ‘reddit’ typed at the end of a million Google searches, a workaround for getting past the listicles to an actual human opinion. That alignment (what we need / what is served) makes Reddit the healthiest of the three, and also the most fragile. If AI slop erodes that trust, there's no getting it back, because you can't astroturf your way to honest.

Across all three examples, the feed is the rent we pay to the platforms via our attention. We didn’t ask for this. In JTBD parlance, we were hired by the platform to pay attention and consume.

We’re all busy people. If the point above is true (feed = a job we didn’t sign up for), why don’t we all quit? Lots of reasons, including an excruciatingly perfect manipulation of human nature, habit inertia, and lack of network capital exchange amongst platforms. In short: retention through switching anxiety, not job performance.

And, honestly, since phones evolved from, well, phones to magic in our pockets, we’ve become increasingly averse to being bored. Standing in line at the deli? We quickly hire Facebook for a dopamine hit.

As I said at the start, I'm trying to understand my own use and to be more intentional about it. Here’s an exercise that takes ten minutes. For each network you use, write down the actual jobs you hire it for. Not the aspirational ones, the real ones: I keep LinkedIn so customers can find me; I keep Facebook for the snowboard instructors group and Marketplace; I keep Reddit for answers I trust. If you can’t name a job, that’s also an answer.

Then change how you enter. The feed is the front door, and this is designed to keep you in the lobby. Go in the side: bookmark the group, the subreddit, the inbox, and navigate straight to them. Do the job, then leave, the way you’d run any other errand. Nobody wanders the hardware store for forty minutes for pleasure (except my wife, but that’s a story for another day).

And judge the time by the job, not the clock. Twenty minutes finding a contractor on a neighborhood group is twenty minutes well spent. Twenty minutes in the feed is a job you aren’t getting paid for.

None of this necessarily means quitting. It’s more about intentional use. The utility layer of these networks is real, and you should use them without guilt or apology. But it's worth being honest about what they can't do anymore: the thing they were originally hired for, real connection among people who share something that matters.

That job hasn’t gone away, it’s just moved on. It lives now in smaller, intentional communities, the ones someone actually designs and tends, where the people in the room choose to be there, and the room exists for them rather than for a sponsor. After twenty-five years of doing this work, I’m convinced these are not a nostalgia act or a niche. They’re where the job gets done now. And in the age of AI, they are only going to get more important.

So keep the directory. Keep the junk drawer. And go find, or build, a room of your own.

It’s hard to think of a word I see more often in the business thought-leader zeitgeist on LinkedIn and Substack than “trust”.

A random sampling from today, collected in 30 seconds:
“You build trust by being trustworthy.”
”Trust is the quiet strength behind every successful product leader.”
”Trust is the foundation of lasting relationships.”
”Agency earns trust, and trust builds connection.”
”Building trust is the true currency of business.”

Obviously, trust as a component of relationships, especially long-term relationships, is important.

But the way we’re talking about trust, as if it were self-evident, singular, and universal, is asking the concept to do too much work. It also creates a recursive hall-of-mirrors effect: trust is defined by reference to trust.

Trust is just… trust?

As Trustocles said to the Athenians, “First, define your terms. Then maybe we can stop trusting trust to explain trust.”

Let’s make this concrete and look at trust in customer relationships. If you focus on “trust” by itself as the way to measure or understand the health of those relationships, you’re going to miss a lot. In fact, trust is one part of a broader concept called Customer Relationship Quality, or CRQ. It’s well-documented, even if it doesn’t get talked about nearly as much.

The Dimensions of a Customer Relationship

In relationship marketing, relationship quality is usually treated as multi-dimensional. The models vary by context, but trust, satisfaction, and commitment recur as central (but not exclusive) dimensions of customer relationship quality.

• Trust is the belief that the other party is reliable, honest, and unlikely to take advantage of the relationship. For a customer, that means believing a brand will do what it says it will do.

• Satisfaction is the customer’s running assessment of the relationship. It is shaped by the gap between what they expected and what they actually experienced.

• Commitment is the customer’s attachment to the relationship and their desire to continue it. This can be practical, emotional, or both.

So, yes, trust matters (obviously). But it is not the whole relationship. Strong customer relationships usually depend on trust, satisfaction, and commitment working together, which is what makes Customer Relationship Quality a more useful frame than trust alone (Rosário & Casaca, 2023).

Why Trust Alone Misses the Bigger Picture

Trust is a useful signal, but it is a limited one. Using trust by itself to understand customer relationships is like driving by looking only in the rearview mirror. It tells you something about what has already happened. It does not tell you enough about where the relationship is going.

Trust can be passive. Commitment is active.

A customer may trust a utility company to keep the lights on, but that does not mean they have a strong relationship with it. It may only mean the company is meeting the minimum expectation. Trust without commitment often means the customer has no urgent reason to leave. Commitment is different. When customers feel attached to a brand, or see a clear practical benefit in staying, they are more likely to keep buying, recommend the brand, and defend the relationship when something goes wrong.

Trust does not guarantee satisfaction.

A customer may trust a software company to keep their data secure and still hate using the product. If the experience is clunky, slow, or frustrating, satisfaction declines. Over time, that dissatisfaction can weaken the relationship, even when the company is still seen as trustworthy. Trust may keep a customer from panicking. It will not necessarily keep them from shopping around.

High trust can make disappointment worse.

Trust also raises the emotional stakes. When a brand customers trust makes a mistake, the reaction can be sharper because the failure feels personal. A less trusted company may annoy people. A trusted company can make people feel betrayed. That is why trust alone is not enough to carry the relationship. Satisfaction and commitment create the buffer that helps a relationship survive mistakes, service failures, and rough patches.

So the issue is not that trust is unimportant, but that trust provides an incomplete picture. If you want to understand the health of a customer relationship, you need to know whether customers are satisfied with the experience and whether they have a reason to continue it.

The higher-stakes problem is that trust can give companies a false sense of security. A trusted brand can still be inconvenient, indifferent, hard to work with, or easy to replace. Customers may believe you will do what you promised and still decide the relationship is no longer worth their time.

The Net

When companies mistake trust as a proxy for relationship health, they miss more actionable signals: declining satisfaction, weaker attachment, and growing openness to alternatives. Satisfaction and commitment may begin to weaken before trust visibly collapses. By the time trust shifts, the relationship may already be in trouble. The relationship signals were likely telling you something earlier. You just weren’t listening for it.

A working map for community leaders and the executives they work with.

When someone asks an AI assistant a question about your product or industry right now, the assistant often reaches into community conversations to answer. Sometimes yours. This has happened quickly, and almost no community team I talk to is operating for it yet.

That is the shift I have been trying to map. Today, I am releasing V4 of the Community AI Framework, along with a companion worksheet called the AI + Community Opportunity Canvas. Both are free. Both are live on a new landing page that will be the canonical home for this work: structure3c.com/community-ai-framework.

The short version of the thesis:
Community is becoming a key ingredient, not a destination. The product team’s recommendation engine, the support team’s agent, the marketing team’s customer story workflow, the executive dashboard’s voice-of-customer summary — all of these are quietly pulling from community. Meanwhile the things community has always been good at, real expertise and real trust between real people, are more valuable now than they were a year ago, because everything else has become abundant and cheap. The work for community leaders is to rebuild the operating model from the member outward and to build the shared language that gets executives into the conversation.

The AI Community Framework

Seven strategic domains and twenty functional areas, covering the full range of where AI is reshaping community work. How community signals get sensed and measured. What AI does for the people in the community. How the member experience changes. How community teams operate. How community translates to business value. Where community feeds long-horizon strategy. How community connects to everything else through agents and APIs.

Each functional area includes key use cases, a stakeholder map, the enabling technologies that matter, and my honest assessment of maturity and confidence. The argument lives in the Notes and Caveats column. The rest is structure.

This is not a buying guide, not a maturity model, not a vendor comparison. It is a map of a domain that is still actively forming, written for people who need to act inside it before it settles.

The AI Community Opportunity Canvas

The Opportunity Canvas is a three-step exercise that takes you from scanning the framework to picking the one bet worth making in the next 90 days. Work through it on your own, with your team, or in a session with the stakeholders the framework helps you identify. From map to move, in a single sitting.

Framework and Canvas History

I built the first version in late 2025, then refined it over a long day in February 2026 with thirty senior community practitioners at the Structure.Community Winter Summit. V4 reflects what held up under that scrutiny and what has shifted since. V5 will ship later this year, sharpened by what people discover when they actually use it.

Start Here

If you only do one thing with the framework, do this: pick one functional area that you are not currently working on, find one primary stakeholder you are not currently talking to, and have a thirty-minute conversation about whether it should be on your roadmap. That is where the value lives. Frameworks do not change anything. Conversations do.

Download both at structure3c.com/community-ai-framework. If you build on it, adapt it, or disagree with something in it, I genuinely want to hear from you: bill.johnston@structure3c.com.

When someone asks an AI tool a question about their business, their product, or their career, the most common next step isn’t to act on the answer. It’s to check it. Pew Research found that 85% of people verify what AI tells them by turning to human sources: peers, colleagues, people who’ve done the work and have opinions grounded in experience.[1]

At the same time, the AI systems themselves are doing something similar. Community forums, expert discussions, and peer-generated knowledge bases are being crawled at unprecedented scale by LLMs because that content represents some of the most reliable, human-validated information available. Earlier this year, I facilitated a discussion with thirty senior community leaders to compare notes on this, and the pattern was consistent: massive volumes of AI crawler traffic across their platforms, with no standard way to measure or account for it.

People go to community to verify what AI tells them. AI goes to community to learn what to tell them. The same infrastructure is serving both sides of the exchange. Most organizations have no idea this is happening, let alone how to manage it.

The missing ingredient in most AI strategies

Pew reported in March that half of Americans say the increased use of AI makes them more concerned than excited, while only 10% say they’re more excited than concerned. About half believe AI will worsen people’s ability to think creatively and form meaningful relationships.[2] Stanford’s 2026 AI Index found that only 33% of Americans expect AI to make their jobs better, well below the global average.[3]

The concern is real, but the more interesting finding comes from Edelman’s research on AI adoption: only 18% of people who resist AI have actually had a bad experience with it.[4] The resistance is anticipatory. People aren’t reacting to what AI got wrong. They’re reacting to the absence of something they need before they’ll act on what AI tells them.

The 2026 Edelman Trust Barometer helps explain what that something is. Seventy percent of global respondents are unwilling or hesitant to trust someone whose values, background, or information sources differ from their own.[5] Trust in institutional leaders has declined. Trust in proximate figures, neighbors, coworkers, direct supervisors, has grown.[5] People are narrowing their circles and deciding who to believe based on accumulated experience with specific people in specific contexts.

This is a pattern I’ve been watching for 25 years in community ecosystems.

The relationship problem underneath the trust problem

The conversation around AI keeps coming back to trust, but trust is what relationships produce over time through accumulated experience between real people. You can’t solve for trust directly any more than you can solve for loyalty by launching a loyalty program. You have to build the conditions that create it. Those conditions are relationships, and for most organizations, the infrastructure for those relationships already exists in their community ecosystems.

I call the specific kind of trust that relationships produce over time relational trust. It’s distinct from brand trust, institutional trust, or the kind of AI reliability that the tech industry means when it says users need to “trust” the model. Relational trust develops through repeated interactions between real people. It compounds. It transfers when someone changes jobs and brings their trusted network with them. And it can’t be manufactured by a marketing budget or generated by an algorithm.

The cross-checking behavior in the Pew data is relational trust in action. The AI provides information. The human turns to people they’ve built relationships with to decide whether the information is worth acting on. Edelman’s research confirms what practitioners in this space have known for a long time: people don’t want to be told what’s good about a new technology from a company’s executives. They need to discover it through personal experience and among peers.[6]

That discovery happens through relationships. The good news is that for the organizations that have invested in community ecosystems, those relationships already exist at scale.

The strategic question

Every organization I work with is investing in AI. The question is whether they’re also investing in the relationship infrastructure that makes AI useful to their customers.

For the companies that have built community ecosystems over years, the answer is already yes, whether they realize it or not. Their customer relationships, their expert networks, their community knowledge bases represent an appreciating asset that becomes more valuable as AI handles more of the transactional layer. AI can generate answers. Community provides the information and the relationships that give those answers usefulness and credibility.

For the companies that are cutting community budgets to fund AI initiatives (or for any reason), the math deserves a second look. You're likely investing in the engine while removing the fuel.

AI deployment can be accelerated with investment. Relationships require investment too, but they also require care and time. The organizations that have spent years building community aren't playing catch-up on AI. They've built the relationship infrastructure that AI depends on.

Endnotes

[1] Bill Johnston. “The Networked Customer in 2026.” Structure3C Substack. January 16, 2026. https://structure3c.substack.com/p/the-networked-customer-in-2026

[2] Pew Research Center. “Key findings about how Americans view artificial intelligence.” March 12, 2026. https://www.pewresearch.org/short-reads/2026/03/12/key-findings-about-how-americans-view-artificial-intelligence/

[3] Stanford Institute for Human-Centered AI. “The 2026 AI Index Report.” April 17, 2026. https://hai.stanford.edu/ai-index/2026-ai-index-report/public-opinion

[4] Richard Edelman via Fortune. “Trust is the missing ingredient in the AI boom.” January 8, 2026. Reporting on the Edelman Trust Barometer Flash Poll: Trust and Artificial Intelligence at a Crossroads (November 2025). https://fortune.com/2025/11/18/trust-missing-ingredient-in-ai-boom-edelman-barometer-poll/

[5] 2026 Edelman Trust Barometer. January 18, 2026. https://www.edelman.com/news-awards/2026-edelman-trust-barometer-society-slides-into-insularity

[6] Edelman Trust Barometer Flash Poll: Trust and Artificial Intelligence at a Crossroads. November 2025. https://www.edelman.com/trust/2025/trust-barometer/flash-poll-trust-artifical-intelligence

On the afternoon of December 4th of last year, I found myself standing in a conference room at the Danish Architecture Center in Copenhagen, working with a small group of people I'd met that morning to map out the future of urban mobility on a flip chart. Urban mobility is not generally in my area of focus, but these were special circumstances.

We were using Post-its to sort forces into three categories: What's pushing the present? What's pulling from the future? What historical weight resists change? The method is called the Futures Triangle, developed by foresight researcher Sohail Inayatullah, and it's a structured way to explore the tensions that shape how the future might actually unfold, or at least to think through it as a rubric.

We were twenty strangers from different countries and industries, mostly from Western Europe, hours into our first day together, and we were already producing collective insight about 2040 that none of us could have generated alone, sitting in front of a screen.

A thought that had been percolating for a while came into much sharper focus.

Twenty-five years of present-tense conversations

I've spent the better part of twenty-five years building community programs and creating community strategy for startups and for large enterprise organizations like Dell, Autodesk, Salesforce, and Siemens. These span a range of strategies, interventions, and programs, including customer advisory boards, innovation programs, community customer ecosystems, and collaborative innovation and design efforts.

The conversations in those (virtual and IRL) rooms, although more often than not remarkable, were very much focused on the present or what was happening in the coming months. Smart, experienced people with deep domain knowledge, sharing what they knew and working on real problems. In part, that was the issue: the focus and framing were on their immediate needs, concerns, and problems, and how to solve them, which is obviously a priority and necessary for survival, but not a ripe substrate for innovation.

But standing in that room in Copenhagen, watching strangers think together about 2040, I could finally see a pattern I’d been sensing for years. Almost every community conversation I’d ever facilitated or observed was anchored in the present tense.

At Dell, IdeaStorm generated thousands of ideas, and the overwhelming majority were incremental improvements to existing products and form factors. At Autodesk, product-specific idea boards are filled with requests for fixes and feature updates. In every community I’ve advised or consulted with, the ideas programs pointed inward: what should we change about what we already make? The infrastructure for gathering collective insight was there, and it consistently got aimed at the smallest possible version of the question.

Nobody was asking these communities to think about the larger context shifts, the emerging trends, or the complex forces reshaping the markets their products were built to serve. I think, in part, because they didn't know how. They didn't have the vocabulary, they didn't have the methodology, and they didn't have a means to position this as an activity worth investing in.

I’d felt that limitation for years without having the right language for it. These programs were very good at gathering what people think about right now. They were almost never designed to explore what might be coming.

The pathway to Copenhagen

Without entirely realizing it, I'd been walking toward the foresight discipline for a while. First, through conversations on the Cohere podcast with people like Ed Finn at Arizona State University and Scott and Susan Smith at Changist, and then, inspired by them, through my own reading and research. In a recent conversation, Ed described imagination as fundamentally "a team sport" and observed that most organizations invest enormous energy training people to stay in their lanes and not ask too many questions, then wonder why nobody saw the change coming. These ideas from Ed and others stuck with me over the last few years. When the opportunity to study applied strategic foresight at the Copenhagen Institute for Futures Studies emerged, I jumped at the chance.

The instructors were clear from the start: foresight is a discipline built on anticipation and preparation, not prediction. The tools help groups of people imagine plausible futures with enough specificity to feel the implications, then work backward to decisions that need to be made now. And the whole methodology is designed to be done collectively, because no individual perspective is wide enough to produce useful futures thinking on its own.

What the method revealed

The collaborative dimension of the methodology really resonated with me, and that's where I dug in. Every exercise in the course over the two days was built for groups. We did some individual work, but we came together as small teams to learn specific methods: the Futures Triangle, pulling out polarities from brainstorming to identify critical uncertainties, building scenario matrices, wind-tunneling strategic initiatives against alternative futures. The quality of the output depends entirely on the range and depth of the people in the room.

Which brought me back to communities.

Twenty people who met that morning in Copenhagen, using a structured methodology, produced four distinct and plausible scenarios for urban mobility in 2040. The work was credible because the method compensated for the fact that we didn’t know each other well. Structure substituted for trust.

Now think about what happens when you deploy the same methodology inside a community of engineers, customers, and partners who have been working together for years within a specific domain. People who already understand the technology, the market dynamics, the regulatory landscape, and, more importantly, different practices, mindsets, and ultimately each other. You wouldn't need the structure to compensate for lack of trust. You'd have both, and the conversation would be grounded in real operational knowledge rather than general expertise.

I kept coming back to this during the flight home. The organizations I've worked with over the years have spent considerable effort building community ecosystems with deep domain knowledge, trust, diversity of perspective, and the kind of relationship infrastructure that makes helpful and honest conversation possible. All of the raw ingredients that foresight methodology depends on. What, until recently, my clients haven't had is the method itself, or the desire to point those communities at the future rather than the present.

Reperception

The foresight discipline has a word for what it produces: reperception. Pierre Wack, one of the pioneers of scenario planning at Royal Dutch Shell, defined it as awakening to the possibility that the future may be different from the past or how you expect it to be.

I think there is a significant opportunity for most organizations to activate large parts of their community ecosystem to do this kind of work. The customers, prospects, partners, and employees within your ecosystem are living in the sheer reality of the market every day. They experience early frictions, behavioral shifts, market trends, and forces. Their reality is comprised of emerging problem sets, solutions, opportunities, and insights that haven’t shown up in any dashboard or analyst report, and won’t for months. In a way, they’re living on the edge of the future. Foresight gives those observations structure. Community gives foresight the one thing it can’t manufacture: a diverse group of people with enough shared context and trust to think honestly about what’s changing, what the implications of that change are, and ultimately how it may play out in the coming months and years.

I’ve spent years making the case that communities are relationship infrastructure, the substrate where ecosystems of customers, partners, and employees form and produce value that can’t be replicated in any other way. These are compounding assets that accrue value over time. I still believe that. But since Copenhagen, I’ve been thinking about a different question: what if the most valuable thing a mature community ecosystem could do is help an organization see what’s coming?

Most executives can describe their community's value in terms of customer success or product feedback. Very few could tell you what their community sees coming. For organizations operating in complex, fast-changing markets, I think that's a blind spot worth examining.

You can find out more about the Copenhagen Institute for Future Studies and their Applied Strategic Foresight class on their website.

——
Structure3C is exploring how foresight methodology and community ecosystems intersect to help organizations anticipate what's coming. If you lead a community, innovation, or ecosystem program at a B2B company and want to explore this territory together, reach out at bill.johnston@structure3c.com.

At last month’s Structure.Community Winter Summit, I watched thirty senior community leaders map their current AI experiments onto a wall. The clustering was striking. Most of the activity concentrated in the same few areas: moderation, search, content drafting, basic analytics. These are useful applications, and several teams are getting real results from them. But they represent a narrow slice of what’s possible.

When we shifted the exercise to “what would you build if constraints were removed,” the ideas expanded dramatically, into business intelligence, personalized member journeys, advocacy discovery, cross-platform listening, and more. The gap is a strategic frame. Without one, teams default to whatever their platform vendor ships or whatever their most technically curious team member experiments with on a Friday afternoon.

I built the Community + AI Framework to address that gap. It organizes the full range of AI opportunities in community into six strategic domains, each with specific functional areas underneath. The goal is to help community leaders see the whole board, identify where they are today, and make deliberate choices about where to invest next.

The framework and a companion pilot design worksheet are available to download at the end of this piece. What follows is a walk-through of each domain and why it matters.

Community Intelligence

Making community signals legible, from individual conversations to ecosystem-wide patterns.

This is the foundation. Every other domain depends on it.

Community intelligence is about making the signals in your community legible, from individual conversations to ecosystem-wide patterns. The functional areas here range from health monitoring and trend detection to network intelligence (mapping who’s actually connected to whom, where the bridges and silos are) to knowledge management, the ability to query years of community discussion history as a structured knowledge asset.

Most teams have some version of basic analytics in place. Platform dashboards show activity metrics, and several vendors shipped AI-powered summaries and recaps in 2025. But the more ambitious applications, like behavioral advocacy mapping or treating community as a queryable knowledge graph, require intentional data architecture and often custom work. The technology is mature enough. The application to community is still emerging.

Member Experience

How members discover, connect, and are guided toward their next meaningful action.

This domain covers how people find content and connections, how the experience adapts to who they are, and how they’re guided toward meaningful next steps.

AI-powered search and content recommendations are shipping now across major platforms, and the improvements are real. One of the more interesting functional areas here is peer and expert matchmaking, using AI to connect members with the right people rather than just the right content. Several teams at the summit described this as a persistent member request that they haven’t been able to address with existing tools.

Member action guidance is another area worth attention. Gamification and badges have been around for years, but AI-driven recommendations (”here are three things you could do this week based on your activity and interests”) require behavioral modeling that most community platforms don’t yet support. The pattern exists in marketing automation. It hasn’t been wired to community signals yet.

Team Augmentation

Helping community teams work faster, know what’s next, build what’s missing, and do it responsibly.

This is where most current AI experimentation lives, and for good reason. AI moderation, content drafting, triage and routing, and workflow automation are the most mature applications in the space. The tooling exists. Adoption is the variable.

But team augmentation extends well beyond operational efficiency. One of the most significant functional areas here is what I’m calling build and experimentation: community teams using AI-assisted development to prototype features, build custom integrations, and close capability gaps where no turnkey product exists. Many of the highest-value opportunities in this framework have no off-the-shelf solution. Community teams that develop the ability to build, even at a basic level, will move faster than those waiting for vendors to catch up.

Content discoverability and AI readiness is becoming urgent. AI search engines are increasingly citing user-generated and social content as authoritative sources. Reddit has been a top-cited domain in AI-generated answers for over a year. And LinkedIn recently surged from around #11 to #5 in ChatGPT’s domain rankings, with posts, articles, and newsletters now accounting for roughly 35% of all LinkedIn citations in AI search responses. The implication for community teams: the content your members create, and the way it’s structured for machine readability, directly shapes whether your community shows up when AI answers questions about your industry, your products, or your competitors. Most community content was never built with this in mind. The teams that audit and optimize for it now will have a significant head start.

Trust, safety, and governance is the final functional area in this domain, and it’s more about process than technology. AI disclosure policies, quality assurance for AI-generated responses, and governance frameworks for AI use in human-relationship contexts are all essential infrastructure that most teams are still writing.

Strategic Alignment

Connecting community activity to business language, executive priorities, and measurable outcomes.

This is where community work gets translated into organizational value.

The functional areas here include translating community activity into strategic signals that align with corporate priorities, and ROI and value translation, the ongoing work of linking community impact to pipeline, retention, and financial metrics. Basic community ROI reporting (deflection, engagement) is available now. The harder work of pipeline and retention attribution requires CRM integration and organizational buy-in that most teams are still pursuing.

The near-future unlock is AI-generated impact narratives. LLMs are well-suited to synthesizing structured community data into stakeholder-ready summaries. But the prerequisite is structured data, which brings us back to the integration challenge that surfaced throughout the summit. The teams that have invested in connecting their data to business systems are the ones whose value story practically tells itself.

One gap the summit exposed in this domain: LLM traffic. Multiple teams are seeing enormous request volumes from AI crawlers consuming their community content. AI systems are treating community-generated knowledge as a trusted source of human-validated information, and that trust was built over years of investment. But no one in the room had a standard way to measure, value, or report it. Content discoverability (in Team Augmentation) addresses the “make it findable” side. What’s missing is the value translation: how do you account for the fact that your community content is training and informing AI systems at scale, and what is that worth? This is an open question the industry hasn’t answered yet, and it belongs squarely in the strategic alignment conversation with leadership.

Innovation and Foresight

Beyond ideation toward shared sensemaking, futures work, and long-horizon exploration.

This is the least explored domain in the framework, and potentially the most valuable for organizations thinking beyond the next quarter.

Innovation and foresight moves past traditional ideation (submit an idea, vote on it) into shared sensemaking and long-horizon exploration. Think community-based futures workshops, high-trust collaboration spaces where customers and partners work alongside product teams, and AI-assisted trend synthesis grounded in real ecosystem signals rather than analyst reports.

I’ve written about the failure patterns of crowdsourced innovation elsewhere in this series. The short version: platforms that treat communities as idea mines consistently fail. The organizations that succeed are the ones that build collaborative infrastructure and relationship-rich environments where problems are shaped, judgment develops, and the community itself becomes a strategic sensing mechanism. AI makes this more powerful by enabling synthesis at a scale previously impossible, but the foundation remains human trust and relationship infrastructure.

Enterprise Integration

The connective tissue that makes community data actionable, where decisions actually get made.

Enterprise integration was the single biggest frustration at the summit. Almost every ambitious idea required connecting community data to systems that community teams don’t control. The barrier is organizational, not technical.

One development worth watching: the Model Context Protocol (MCP), an open standard for connecting AI systems to data sources, was adopted by major AI providers and donated to the Linux Foundation in late 2025. It’s becoming an industry standard, and community-specific MCP servers are an emerging opportunity. The ability to pull community data into AI-powered analytical and operational workflows, without building custom integrations from scratch, could meaningfully change the integration picture over the next year.

From framework to action

A framework is only useful if it leads to decisions. At the summit, I introduced a companion worksheet designed to move community leaders from orientation through opportunity identification to a concrete pilot plan.

The worksheet works in three stages. First, you read through the six domains and their functional areas and notice where you feel resonance, tension, or surprise. If something makes you think “we should be doing that” or “we tried that and it was a mess,” pay attention to that reaction. Second, you identify five to seven ideas and then filter down to the one or two with the strongest pull, considering where the pain is sharpest, what would benefit members most, where impact would be most visible, and what’s feasible in 90 days. Third, you design a pilot: an elevator pitch for what you’re testing, who benefits, what you need, and how you’ll know if it’s working.

The worksheet includes a specific format for the pilot pitch: “We will test [what you’re building or changing] with [who specifically] to learn whether [the question you’re trying to answer] within [timeframe].” That specificity matters. It’s the difference between “we should do something with AI” and a plan you can start on Monday.

Both the framework and the worksheet are available for download here (short registration). The best way to say “Thank You” for access is to like and share this post!

Thanks to the members of the Structure.Community network for giving feedback on early versions of the worksheet and framework, especially Chris Catania from ESRI.

The Community + AI Framework (V3) and the AI Pilot Design Worksheet are free to use and share. If you’d like to discuss how they apply to your specific program, reach out at bill.johnston@structure3c.com.

Last Wednesday, thirty senior community leaders traveled to San Francisco, cleared their calendars, and spent a full day discussing two questions: how is AI changing community practice, and how do community leaders build a business case that reflects what community actually does? They chose to show up in-person, to be in a room with people who understand what they do, because in moments of massive change, the best source of information is the community of leaders and practitioners actually doing the work.

That choice tells you something about the value of community.

I opened the day with five provocations and asked the room to pressure-test them. The room confirmed all five and added texture and detail I wouldn’t have gotten on my own. What follows is what I think the executives who fund and evaluate community programs need to hear right now.

Choosing to practice takes courage right now

This is a hard time to lead community. Budgets are being cut. Roles are being eliminated. Platforms get acquired, and the roadmap changes overnight. The headlines about declining engagement on sites like Stack Exchange are reaching the C-suite, and some leaders are concluding that community is obsolete.

The people in that room are choosing to practice anyway. And the distinction matters because those who stay through this transition will define what community becomes next.

Our discussion confirmed the “community is dead” narrative is a leadership communication problem, not a reality. Community’s surface-level engagement patterns are changing as AI handles transactional queries. The underlying trust, expertise, and first-party data that make community valuable are more relevant than ever. But communicating that requires new language and new metrics, and most teams are still developing both.

The community and community-skilled leaders who are succeeding right now are the ones who have stopped defending the old story and started telling a new one. They’re reframing community from a destination people visit to an ingredient that mixes into every customer-facing function. One team shared that their community content, which used to generate 7 million visits a year, now produces 209 million impressions because it’s distributed across products, search, AI interfaces, and partner experiences. Same content but in a different frame led to a dramatically different reception from leadership.

The social exodus is our opportunity

Social media usage peaked in 2022. People are leaving mass social spaces, or using them significantly less, and looking for something more private, more contextual, more meaningful. Community teams have been building exactly that for years.

The room reflected this, as multiple leaders described growing engagement in smaller, more intentional spaces even as broader platform metrics soften. The pattern is consistent with what I wrote about in my last piece: the same impulse that’s driving people away from industrialized social media is driving them toward the kinds of spaces community teams know how to build, that are intimate, purposeful and led by the people who participate.

This should be obvious: as trust in mass platforms erodes, the organizations that have invested in genuine community relationships hold an asset that compounds over time and becomes harder to replicate the longer it runs. The question is whether leadership recognizes that asset before it gets impacted in the next budget cycle.

Your community already has an AI strategy, whether you created one or not

I told the room: your community has an AI strategy right now. It’s just implicit. Platform providers are making decisions about AI integration. Automation is being layered in. Corporate AI policies are shaping what’s allowed. Members are showing up with AI-assisted contributions, and in some cases AI-generated ones. All of that is happening whether or not anyone on the community team has been consulted.

The room’s response was immediate recognition. Two leaders opened with framing presentations on how they’re positioning community and AI with their own leadership. Then the room mapped what they’re actually doing today, what they’d build if constraints were removed, and refined the most promising ideas into pilots they could start within weeks.

The most compelling cases that surfaced move beyond automation and into richer and more connected human experiences. AI translation connecting forums across languages, expanding the pool of potential answerers and increasing solution rates by up to 50%. AI-powered discovery surfacing advocates that teams didn’t know existed, growing advocacy cohorts by 5x. AI turning raw community conversations into structured product intelligence reports that product teams actually read.

Every example followed the same logic: AI handles scale and friction, humans provide trust, judgment, and authentic connection. And every pilot that people committed to testing included explicit human oversight. No one in the room proposed fully autonomous AI operations. Community work is fundamentally about human relationships, and these leaders aren’t willing to hand that over.

One practitioner framed the opportunity with a line I keep coming back to: “intelligence is a commodity, connection is not.” If your community strategy treats AI as a replacement for human connection, you’ve misread the moment. If it treats AI as an accelerator for human connection, you’re building something durable.

Community teams are sitting on a strategic asset, and most organizations don’t know it

I called this the Community Data Paradox. Community is a type of relationship infrastructure. The network of customer relationships that a company invests in over time, and that accrues value the longer it operates. Those relationships generate some of the most valuable first-party data a company owns: conversations, behaviors, sentiment, intent signals, and trust. And in most organizations, that data lives in complete isolation from the systems where business decisions get made.

This was the single biggest frustration in the room. Almost every future-state idea required connecting community data to CRM, product analytics, support platforms, training content, or marketing automation. Very few teams have those integrations in place. The barrier isn’t technical imagination. It’s organizational: budget, prioritization, and the difficulty of getting community onto the enterprise integration roadmap.

The teams that are demonstrating value most effectively have invested in data partnerships with CRM, product, and finance teams. They can show retention differentials between accounts with active community members and those without. They can connect community engagement to pipeline influence using the same attribution models marketing already uses. They can demonstrate that advocacy community members complete learning courses at a 15% higher rate than the general population, a direct, measurable impact on an outcome that the business already tracks.

The teams that are struggling are the ones whose community data lives in isolation. The measurement problem, the attribution challenge, the inability to demonstrate full business impact: it all traces back to the same root. Every team in the room that was succeeding with leadership had one thing in common: their community data was connected to the systems where business decisions get made.

There’s a wild card here too. Multiple teams are seeing enormous request volumes from AI crawlers accessing their community content. LLMs are treating community-generated knowledge as a trusted source of human-validated information. That trust was built over years of community investment. Whether and how to account for this in community’s value story is an open question the industry hasn’t addressed yet. But the machines have already figured out what some executives haven’t: authentic community content is among the most valuable data assets an organization can produce.

The value story needs a reboot

If we pull on the threads from the preceding provocations, we see a clear pattern emerging: that community creates value for every customer and partner-facing function, from product, support, marketing, sales, customer success, and training. But most community teams are still telling their story to one stakeholder in one language, which is usually support, usually focused on cost reduction.

The afternoon session took this on. The room mapped how they currently demonstrate value, and the honest assessment was sobering. Most stakeholder reporting is surface-level, fragmented, and burdensome to produce. When I asked who produces a dashboard or report that the C-suite reviews at least quarterly, only a few hands went up.

The attribution conversation was the liveliest exchange of the day. One leader shared that his team had shown community touching $8 billion in revenue because 90% of customers used it. Then the community was taken away, and revenue went up because the company raised prices. His point: be careful with attribution claims you can’t defend under pressure.

But another pushed back on the double standard. Marketing teams routinely claim attribution for someone reading a blog post three months before a purchase. If that’s the accepted model, community should use it too. Hedge toward underestimation so you can defend the numbers. But don’t distance yourself from revenue entirely. Revenue is the only language most business leaders understand.

The room’s consensus was clear: community attribution should be honest and defensible, but not timid. Use the same attribution model your organization already accepts for other departments. Frame community as associated or influenced rather than directly caused. But don’t cede the revenue conversation to avoid the discomfort of imperfect measurement.

The deeper issue is that the value story itself needs to evolve. Deflection metrics made sense for a time, when community was positioned primarily as a support channel. They don’t capture what community has become, and are holding the value story back. The teams making progress are the ones who have reframed community as relationship infrastructure, an investment in the network of customer relationships that supports retention, expansion, product intelligence, advocacy, and increasingly, the quality of data that AI systems depend on. That framing changes the conversation from “what does community cost?” to “what does this network of relationships produce?”

The room was proof

I ended the day the way I started it. Thirty people chose to be in a room together because they wanted to think with other humans about problems that matter to them. They shared real numbers, real failures, and real experiments with the kind of candor that only happens when trust is present. Several committed to specific pilots they planned to start within days.

That’s the case for community, right there. The fact is that when the questions are hard enough, and the stakes are real enough, people still seek out other people.

AI is changing the surface of community work. Fewer visits, fewer questions, softer engagement numbers. At the same time, it’s making what community actually produces, authentic expertise, trust signals, first-party intelligence, real human relationships, more strategically relevant than it has ever been.

The leaders who showed up that Wednesday are navigating real organizational resistance, volatile budgets, and the persistent difficulty of proving what they know to be true. They’re doing it because community is a vocation for them, and because they understand that the relationships their teams steward are among the most durable assets a company can build.

The executives who understand this know it because a strong community leader helped them see it. The ones still cutting these programs are telling you something about how they think about customers. Not as a network of relationships that compounds in value over time, but as a cost to be managed. That distinction reveals more about an organization’s leadership than any earnings call.

Structure.Community brings senior community leaders together for creative retreats to explore the questions that matter most. The next gathering is this fall. If you lead a community or ecosystem program at a B2B company and want to be part of the conversation, reach out at bill.johnston@structure3c.com.

A recent trip back to Kentucky gave me time to think about how we connect online, how that’s changed over the years, and where it might be heading. A visit to the Wendell Berry Center in New Castle added some additional texture to those thoughts. More on Berry in a moment, but let’s start in Chicago, circa 1978.

In February 1978, Chicago was buried under the Great Blizzard. Ward Christensen and Randy Suess, two computer hobbyists who couldn’t get to their club meeting, decided to build something instead. Using a spare S-100 bus computer and an early modem, they created the Computerized Bulletin Board System, the first BBS. It went online on February 16th, ran on a single phone line, and connected 253,301 callers before it was finally retired. That was nearly 50 years ago.

They built it because they were snowed in and wanted to talk to their friends.

That impulse of people building something together because they sought connection is where online community begins. And for about fifteen years, that’s largely how it stayed. Not because anyone had a philosophy about it, but because the technology imposed constraints that, by chance, produced (mostly) healthy communities.

The Electronic Homestead Era

Because long-distance phone calls were expensive, most BBSes were local. The person you argued with about Star Wars at midnight might just be coaching your kid’s soccer team on Saturday. Sysops, the people who ran these systems, usually from a spare computer in a bedroom, hosted meetups at parks and pizza joints. Reputation was real because most users were local. At their peak in the mid-1990s, there were more than 60,000 BBSes serving approximately 17 million regular users.

The WELL, the Whole Earth ‘Lectronic Link, launched in 1985 out of the houseboat community in Sausalito, California. Stewart Brand, who’d created the Whole Earth Catalog, and Larry Brilliant set it up with twelve phone lines and a VAX computer. Its founding policy was four words: “You Own Your Own Words.” Members were accountable for what they said. The Electronic Frontier Foundation and Craigslist both emerged from people who met there. Howard Rheingold wrote The Virtual Community based on his experience as a member.

Usenet, which predated the web by over a decade, developed its own self-governance. In 1987, the community undertook the “Great Renaming,” a massive reorganization into hierarchical categories, debated and decided by the people who used the system. It was messy and contentious, as democratic processes tend to be. But the people affected were the ones deciding. Norms were enforced through social pressure, FAQs, and the collective invention of “netiquette,” a word that now sounds quaint but represented a genuine community effort to figure out how to behave well in a new kind of space.

These weren’t utopias. BBSes could be cliquish, exclusionary, and overwhelmingly homogeneous. Sysops wielded unilateral power. Access required equipment, resources, and knowledge that many people lacked. But the fundamental structure was sound: small enough that reputation mattered, participant-governed, and organized around shared interests rather than commercial data extraction.

The Electric Fences

Then came the electric fences.

In September 1993, AOL opened its floodgates to Usenet. The annual trickle of university freshmen who needed to learn the norms became a permanent deluge. Usenet veterans coined the term “Eternal September,” the month that never ended, when the self-policing culture was simply overwhelmed by volume. The community’s immune system couldn’t scale.

What followed over the next two decades was a gradual but thorough industrialization of both corporate online communities and social media at scale. The shift wasn't dramatic. It occurred as factory farming did, through a series of individually rational decisions that produced cascading unintended consequences.

Companies discovered that customers in a forum would answer each other's questions, and someone quickly did the math: a support phone call costs roughly $100, while a peer-answered forum post costs essentially nothing. "Community" became a line item in the customer support budget, and the vocabulary shifted accordingly: call avoidance, case deflection, self-service resolution. Meanwhile, mass social media platforms discovered that outrage and anxiety were more engaging than connection, and optimized accordingly. In both cases, the people who created the value were not the ones who captured it.

Anyone who has spent time in this industry, as I have for twenty-five years, has watched the center of gravity make an unfortunate shift from "how do we create genuine value for members?" to "how do we prove ROI through deflection metrics?"

Interlude: A Visit to the Wendell Berry Center

A couple of months ago, I drove to New Castle, Kentucky, to visit The Berry Center. New Castle is the kind of small county seat that exists all over the American interior: a courthouse square, a few storefronts, the quiet that comes after decades of economic drift. The Center is just off the square, in a modest building you could walk past without noticing.

Inside, the first floor is essentially a working library. Shelves of Berry’s writing, some displays about his work and his family’s, including his father’s efforts on tobacco policy in the first half of the last century, are a reminder that the Berrys have been thinking about the relationship between land, economy, and community for generations.

One of the Center's directors walked us through the space and talked about their local programs, which focus on the practical, unglamorous work of helping people rebuild connections to the land and to each other.

Then I noticed a display case near one wall. Inside were two early copies of the Whole Earth Catalog. I hadn’t expected that. Stewart Brand, who created the Catalog, went on to co-found the WELL in 1985, the community that became the template for everything that followed. The thread ran directly from Berry’s agrarian philosophy through Brand’s catalog and into the first serious attempt to build an intentional online community. It was a full-circle moment for me to realize that the early community builders on the West Coast who inspired me were, in turn, inspired by agrarian community builders and advocates like Berry in my home state of Kentucky.

Something Berry wrote kept coming back to me as I drove home. In a poem called XI, he imagines speaking across time to those who will inherit the land after the present ruin, and asks them simply to come with the intention of staying. Not to extract or optimize, but to stay, and to “treat it well.”

That’s the thing we forgot how to do with community.

But something is shifting. Mass social media participation has been declining since 2022. Just as many people have developed a growing aversion to factory farming and ultra-processed food, they are pushing back on the industrialized, ultra-processed social media experience.

Green Shoots Springing Up in the Cracks

In 2022, Elon Musk’s acquisition of Twitter demonstrated what happens when a single owner controls the town square. The mass migration to Mastodon and Bluesky wasn’t just platform switching. It was people seeking community structures where no single entity could unilaterally change the rules or impose their own weird monoculture.

In 2023, Reddit’s unpaid moderators, the people who had built and maintained communities generating billions in value, went dark in protest when the platform moved to monetize the infrastructure those volunteers had created. It was the community equivalent of farmers losing control of their seeds to an agribusiness conglomerate. And this isn’t the first time volunteer moderators have revolted - AOL settled a lawsuit with volunteer moderators for $15 million back in 2010. History doesn’t repeat, but it rhymes.

The fediverse, the federated network of independent servers running Mastodon and other platforms, now encompasses over 15 million users across 22,000 servers. Each server sets its own rules. They connect by choice. When Meta’s Threads began integrating with the network, over 800 servers preemptively blocked it. Community self-determination in action. Bluesky’s AT Protocol ensures that users can carry their identity and social connections to another server if they don’t like how things are run. That’s the digital equivalent of seed sovereignty (the right of farmers and communities to save, use, exchange, and sell their own seeds).

Meanwhile, small intentional community spaces are quietly proliferating: Discord servers, private Slack groups, group chats, niche forums. They function a lot like BBSes did: local in affinity if not geography, governed by people who are also participants, small enough that reputation matters. Nobody calls them a movement. They’re just people rebuilding what was lost, the same way backyard gardeners and farmers’ market vendors are quietly rebuilding local food systems without waiting for permission from the industrial complex.

To be clear: I don’t think the answer is to go back to 1985. I don’t think the answer is to brand what’s emerging with a catchy name and call it a movement (looking at you, Cozy Web). I think the answer is simpler and harder than that: we have to stop treating community as a cost center to be optimized and start treating it as what it actually is, relationship infrastructure that creates value for everyone involved, not just the organization that hosts the platform.

The organic food movement didn’t replace industrial agriculture. It grew alongside it, offering an alternative for people who wanted something different and were willing to invest in it. I think that is what’s happening with community too. The factory farms aren’t going away. But the people who tend gardens are finding each other again.

I’ve spent most of my career in the space between companies and their customers. Building communities, designing collaboration programs, and advocating for investing in long-term relationships.

So I know what it looks like when it goes wrong.

Someone gets excited about “engaging customers in innovation.” They find a budget, pick a platform, and sound an open (and unstructured) call for submissions. Ideas pour in. Nobody inside has the authority or capacity to act on them. Contributors get silence instead of feedback. Participation drops. The program gets deprioritized. Eventually, someone writes a case study about “lessons learned” that hardly anyone reads.

In my previous article, I explored the patterns behind this cycle: the Idea Void, the Vertical Integration Trap, the Ideation vs. Realization Gap, the Engagement ≠ Value Confusion, and the emerging Dead Internet Problem. They share a root cause: treating participants as a source of free labor rather than investing in the relationships and infrastructure that enable collaboration.

This article offers a framework for building programs that produce outcomes, or diagnosing why your existing ones aren't.

One language note before we get into it. I will use the term "collaborative innovation infrastructure" instead of "community" throughout this piece. Not because "community" is a bad word (I've built my career around it), but because it has become a catch-all phrase and, when used without context, lacks shared meaning. An open idea portal is not the same as a customer advisory board, which is not the same as a developer ecosystem, which is not the same as a six-week futures program in which executives explore the hydrogen economy. Calling them all "community" lets people file the whole category under “discretionary customer engagement stuff.” I want to be specific about what we're building and why.

The Shift From Extraction to Infrastructure

The extraction model treats customer relationships like vending machines. Insert challenge, receive ideas. This type of “relationship” (such as it is) is transactional, episodic, and, now that AI can generate ideas for free, increasingly pointless.

Infrastructure is different. It’s an investment in relationship-building capability that compounds over time. The bonds between participants and hosts strengthen. The institutional knowledge deepens. The question shifts from “what ideas can we extract?” to “what innovative work do these relationships make possible that we couldn’t do alone?”

The Five Layers of Collaborative Infrastructure

Organizations that get this right tend to address five layers. Skip one and the problems show up downstream, usually disguised as “engagement issues” or “the community isn’t delivering value.”

Fit: Should We Do This at All?

Most programs skip this layer entirely. Someone had an idea, found budget, and started building. The strategic rationale was later retrofitted, if at all.

Fit asks two things: Does this belong in our innovation portfolio? And can we actually absorb what we’d receive?

Different innovation objectives require fundamentally different approaches. Incremental improvement, adjacent expansion, and breakthrough innovation need different levels of openness, different participants, and different tolerance for ambiguity. Running a breakthrough problem through an incremental mechanism produces frustration on all sides. Meanwhile, who owns outcomes? What downstream resources are pre-committed? Who has the authority to actually approve, block, or act on what comes in? Programs don't fail from lack of ideas. They fail because no one inside is set up to do anything with them.

Frame: What Are We Actually Building?

Frame answers the structural questions. What are we asking people to do? Who participates? How will they engage? What are the rules?

Most of this comes down to specificity. Vague prompts produce vague submissions, then internal teams spend more time decoding what contributors meant than it would have taken to solve the problem themselves. Clear boundaries matter too. LEGO Ideas works because the boundary is obvious: fans design, LEGO manufactures. Quirky failed because the boundary blurred. (Try running injection mold tolerances past a thousand enthusiastic strangers.) The engagement format must match the purpose. Always-on platforms for continuous input. Time-bound challenges for focused problem-solving. Cohort programs for co-development. Live convenings for relationship-building. The failure is using one format for everything.

Exchange: Why Would the Right People Show Up?

If Fit and Frame are about organizational design, Exchange is about people. What makes the right participants contribute and stay engaged?

Every program needs a clear answer to: what do participants get that they can’t get elsewhere? Access to peers they wouldn’t otherwise meet. Influence on product direction. Early insight into strategy. Recognition. Compensation. “Exposure” is not a value proposition. Neither is “being part of a community.” The other structural failure is over-rewarding ideation and under-rewarding everything else. Testing, critique, documentation, synthesis... if you only celebrate idea generation, that’s all you’ll get. And passion will not compensate for structural extraction. At least not for long, and not from the people you most need.

Ops: How Does This Actually Run?

Ops is the least glamorous but most important layer. Where do contributions land, and who is accountable for doing something with them?

Salesforce’s IdeaExchange assigns product managers to specific categories. Ideas don’t land in a generic queue. They land with someone who has the authority to act. Adobe replaced its always-on platform with two focused cycles per year. Constraints create focus. The most important operational commitment is feedback. Not “thanks for your submission” but “here’s what we did with it, or here’s why we didn’t.” I’ve seen silence destroy more programs than any other single factor. And with AI now part of every conversation, you need to be explicit about what humans provide (judgment, context, novel framing) versus what AI contributes (synthesis, screening, pattern-matching). If you can’t articulate why you need humans in the loop, you’ll either waste their time or erode their trust.

Yield: Is This Actually Working?

Yield is the accountability layer. After all the design and operation, what did we actually get?

Implementation rate is one key indicator. Toyota’s famed internal suggestion system achieves 70% implementation, which is why it’s operated for decades while idea platforms come and go. If contributions aren’t being acted on, participants notice. They leave. Beyond implementation, connect participant input to actual business results: revenue generated, costs avoided, risks mitigated, time saved. This is harder than counting submissions, which is why most programs don’t do it. Institutional memory is also an asset. What does the organization retain if the program owner leaves? What do returning participants know that new ones don’t? If every engagement starts from zero, you’re not building infrastructure; you're really just running events.

Suggestions for Using the Framework

As a design sequence: Work through Fit → Frame → Exchange → Ops → Yield when building something new. Remember that each layer is dependent on the preceding layers.

As a diagnostic: Audit existing programs layer by layer. Most dysfunction stems from skipped layers, typically Fit (no strategic clarity) or Exchange (incentive mismatch). What looks like an engagement problem is often a Fit problem in disguise.

What’s Next

The next article in this series will introduce exercises designed to allow you to evaluate your programs with each layer of this framework. Test the framework in your own context before you commit real resources to it.

Most organizations already have the raw ingredients for collaborative innovation: Motivated customers, willing partners, and real problems worth solving together.

What's missing is the structure to make it work.

Interested in discussing collaborative innovation and how to build more value through customer relationships? Join Structure.Community. Next in-person event: February 25th, 2026, San Francisco.

Between 2010 and 2022, some of the most celebrated crowdsourcing ventures collapsed.

Quirky raised $180 million to become an “invention machine,” crowdsourcing everything from product ideas to branding. At its peak, it had over a million community members submitting ideas for consumer gadgets. It filed for bankruptcy in 2015.

Local Motors promised to revolutionize automotive manufacturing with “microfactories” and co-created vehicles. Its community designed the Rally Fighter and the Olli autonomous shuttle. The company shut down in 2022.

Corporate idea portals told a similar story. MyStarbucksIdea, Nokia’s IdeasProject, Dell’s IdeaStorm: all launched between 2007 and 2008, all drew millions of participants and thousands of ideas… and were eventually discontinued.

These weren’t fringe experiments. They were well-funded, widely praised, and put forward as proof that the “wisdom of crowds” could transform innovation. Harvard Business School wrote case studies about them. FastCompany featured them as cutting-edge examples. Conference keynotes celebrated them.

And they all failed.

With the benefit of hindsight (and tracking the impact of generative AI), five patterns begin to emerge. The bad news is that they are probably active in your organization right now, whether you’re running an internal idea platform, a customer co-creation program, or a partner innovation ecosystem.

But before we examine what went wrong, we need to address something uncomfortable about what supposedly went right.

The Survivorship Bias Problem

When executives ask “what does good look like?”, experts trot out dated case studies, including LEGO Ideas, Threadless, and InnoCentive.

These are fine examples. They’re also 15-20 years old, frequently cited without rigorous evidence, and can be potentially misleading, or at the very least, impossible to replicate.

• LEGO Ideas is real, but how much of LEGO’s $10.5 billion (DKK 74 billion) in 2024 revenue (the most current year reported) is actually attributable to the Ideas program versus the broader brand, licensing deals, and product portfolio? The data is murkier than the case studies suggest. The 10,000-vote threshold filters ruthlessly; the vast majority of submissions never reach review. And the 1% royalty to creators, while meaningful recognition, isn’t exactly a partnership model.

• InnoCentive (now part of Wazoku) demonstrated something important: solvers from outside a problem domain often outperform specialists. But dig deeper, and you find that open crowdsourcing solved only one-third of the problems that internal R&D couldn’t crack. Two-thirds remained unsolved even with access to a global network of 400,000 scientists. The platform was acquired in 2018, suggesting the standalone economics were challenging.

• Threadless pioneered crowd-designed products and built a community of 3 million. But those are peak metrics from the early 2010s. The company has pivoted significantly since, suggesting the original model required substantial evolution.

These examples describe what worked in particular conditions at a particular time. They're historical artifacts to study, not templates for modern innovation programs.

Quirky, Local Motors, and Nokia offer something more useful: clear evidence of what breaks. These weren't random failures. They reveal five structural patterns, patterns that are probably active in your organization right now, whether you're running an internal idea platform, a customer co-creation program, or a partner innovation ecosystem.

The Patterns

The Four Failure Patterns

1. The Idea Void: Thousands submitted, almost none implemented, contributors abandon ship

2. The Vertical Integration Trap: Crowdsourcing ideas while trying to internalize everything else

3. The Ideation vs. Realization Gap: Applying crowd models to industries that require massive infrastructure

4. The Engagement ≠ Value Confusion: High participation metrics, zero business impact

And One Emerging Risk

1. The Dead Internet Problem: AI-generated submissions drowning out human signal

Failure Pattern #1: The Idea Void

What it looks like: Thousands of ideas submitted, single-digit implementation rates, no feedback to contributors, declining participation over time.

Why it happens: Organizations launch idea platforms because the technology is easy, not because they have a pathway to implementation. The ideas land in a queue reviewed by a team with no product authority, no budget, and no incentive to champion external suggestions. Research confirms the dynamic: when firms receive too many ideas, they focus on familiar ones, defeating crowdsourcing’s purpose entirely.

What actually happened: Dell’s IdeaStorm nearly died of this disease.

Launched in 2007 as a bold experiment, IdeaStorm was one of the first enterprise idea platforms. It drew nearly 15,000 suggestions in its first few years and built a community that once numbered over a million members. Harvard wrote about it. Conferences celebrated it.

By 2010, Dell’s attention had moved to other projects. One community manager served that million-member constituency. Suggestions piled up and gathered dust. Contributors noticed. The community was on the verge of revolt.

The fix wasn’t a new platform. It was rejoining the conversation Dell had started. The turnaround required “Idea Partners” to route suggestions to actual decision-makers across the 110,000-employee organization. “Storm Sessions” created real-time dialogue on specific topics. And critically, the program hired its most vocal critic, a 911 dispatch supervisor from Florida who had been publicly challenging Dell to participate, as the community manager responsible for bridging the customer-company gap.

IdeaStorm was eventually retired in 2018, but on its own terms, not as an abandoned ghost town. The lesson: idea platforms die when companies stop showing up.

Meanwhile, Toyota achieves approximately 70% implementation of employee suggestions. In 2023, that meant roughly 570,000 ideas actually became operational improvements. The difference isn’t volume. It’s organizational commitment to closing the loop.

Diagnostic question: What percentage of ideas submitted to your platform in the last 24 months have been implemented? If you don’t know the number, that’s your answer.

Failure Pattern #2: The Vertical Integration Trap

What it looks like: Crowdsourcing the “fuzzy front end” (ideation) while trying to internalize the “hard back end” (development, manufacturing, quality control, distribution). The result is spiraling costs, quality problems, and strategic incoherence.

Why it happens: The fantasy of the “invention machine,” a system that turns crowd ideas directly into market-ready products, ignores where value (and supply) chains actually break down. Communities are excellent at generating ideas but structurally incapable of assessing things like manufacturing feasibility, regulatory requirements, or unit economics.

What actually happened: Quirky tried to manufacture hundreds of distinct product categories, from air conditioners to power strips. Each category required different supply chains, quality standards, and go-to-market approaches. The community could vote on what seemed cool. It had no way to evaluate what was buildable and profitable. Quirky ended up trying to be good at everything and profitable at nothing.

Contrast this with LEGO, where the community submits designs, but manufacturing and quality control remain strictly internal. The community’s role is bounded. LEGO doesn’t ask fans to optimize injection molding.

Diagnostic question: Where exactly does community input stop and organizational capability take over? If that boundary isn’t clear, you’re asking the crowd to do things it structurally can’t.

Failure Pattern #3: The Ideation vs. Realization Gap

What it looks like: Applying crowd innovation models to domains where the distance between “idea” and “implementable product” requires massive capital investment, regulatory navigation, or specialized infrastructure.

Why it happens: “Democratizing innovation” is an appealing narrative. But some industries have structural barriers that no amount of community enthusiasm can overcome. A thousand passionate contributors don’t add up to one crash-test facility.

What actually happened: Local Motors built gorgeous vehicles. The Rally Fighter looked like nothing else on the road. The Olli autonomous shuttle was genuinely innovative. But developing road-legal vehicles requires enormous investment in safety testing and certification. The “microfactory” model, while agile, couldn’t achieve economies of scale. And the community, however talented, wasn’t large enough to create a meaningful market.

Hardware is hard. Open innovation in capital-intensive industries requires an anchor partner with deep pockets and regulatory expertise. Standalone open hardware ventures face a nearly insurmountable “valley of death” between prototype and production.

Diagnostic question: What’s the realistic capital and capability gap between a community-generated concept and a shippable product in your domain?

Failure Pattern #4: The Engagement ≠ Value Confusion

What it looks like: High participation metrics (submissions, votes, comments) with no corresponding business impact. Leadership eventually notices and pulls funding.

Why it happens: Community programs often inherit measurement frameworks from marketing: engagement metrics, participation rates, NPS scores. These measure activity, not outcomes. You can have a highly engaged community that contributes nothing to innovation, and a small, quiet group that transforms your product roadmap.

What actually happened: Pepsi’s Refresh Project generated millions of votes for community improvement ideas. It was a social media success. It was also a business failure; Pepsi actually lost market share during the campaign (reportedly ~$350 million in value). The engagement was real; the connection to why people buy soft drinks was nonexistent.

The program was cancelled despite its popularity metrics. Leadership eventually asked the right question: “What did this produce?”

Diagnostic question: If your innovation community disappeared tomorrow, what specific business outcome would degrade? Can you name it precisely?

Emerging Risk: The Dead Internet Problem

The first four patterns have already killed programs. This one is still gathering force, but the implications are obvious.

What it will look like: Declining signal-to-noise ratio. Increasing volume of low-quality, generic, AI-generated submissions. Filtering costs exceeding the value of the ideas being filtered.

Why it will happen: Generative AI has made producing “ideas” essentially free. If your platform accepts open submissions without verification, you’re increasingly processing synthetic content.

What is actually happening: This pattern is still emerging, but the indicators are visible. AI-generated content floods social platforms, attracting engagement from other AI systems and creating a “Potemkin village” of activity. For innovation communities, the implication is stark: if your next Idea Jam receives 10,000 submissions and 9,000 are ChatGPT-generated, the cost of identifying the 1,000 human contributions exceeds the value of running the program.

The solution isn’t banning AI assistance; that’s unenforceable and potentially counterproductive. The real question is what you’re sourcing from humans (judgment, context, lived experience) versus what AI can legitimately contribute (synthesis, iteration, refinement).

Diagnostic question: How would you know if 30% of your community submissions were AI-generated? Do you have any verification mechanism, or are you assuming good faith at scale?

The Pattern Beneath the Patterns

The five patterns share a root cause: treating the community as a source of free inputs rather than as infrastructure for building relationships and capability.

The extraction mindset (”how do we get more ideas from more people at lower cost?”) leads directly to idea voids, engagement theater, and contributor fatigue. It treats participants as a resource to be mined rather than as partners in a shared enterprise.

The organizations getting this right have made a different bet. They stopped asking “how do we crowdsource innovation?” and started asking “how does our community ecosystem strengthen our innovation capability across the entire portfolio?”

What This Means for Your Organization

If you recognized your program in any of these failure patterns, you’re not alone. Most corporate innovation communities are still running some version of the 2007 playbook. Dell IdeaStorm and MyStarbucksIdea were retired years ago, but their architecture persists because it’s what the platforms support and what the case studies describe.

These failures are structural, which means they’re fixable. But better software or more gamification won’t do it. What’s required is a different conception of what community-based innovation is actually for.

The question isn’t “how do we get more ideas from more people?”

The better question: how do we position community as a force multiplier across our entire innovation portfolio, from early exploration through validation, development, and ecosystem growth?

That’s where I’m headed with the next article.

Interested in discussing collaborative innovation, the evolving role of customer communities, and how to create more value through customer relationships with a select group of brilliant CX, Community, and Digital leaders?

Join Structure.Community. Our next in-person event is on February 25th, 2026, in San Francisco.

A decade ago, I argued that customers don’t make decisions in isolation. We live and work in networks, shaped by peers, reinforced by shared experience, and validated by other people like us. The customer journey has always been messy and connected, no matter how much we prefer linear diagrams and neat funnels.

This “networked customer” thesis has held up, but the network itself is evolving.

In the past few years, two changes have altered how customers gather information and make decisions. First, mass social media has stopped functioning as a shared, reliable layer of connection. The platforms that once served as the default network layer have become cacophonous spaces with plummeting engagement. Second, AI tools have entered the decision-making process as new participants. Customers have started using them for their initial research.

The network hasn’t disappeared. It is in the process of fragmenting and reorganizing. Understanding that reorganization is critical.

The AI Paradox: High Adoption, Low Trust

Here is the tension: AI adoption is accelerating while AI output isn’t fully trusted.

The adoption numbers are clear. Sixty-one percent of Americans report using AI in some form, and 37% now start their online searches with an AI tool rather than a traditional search engine.¹ Nearly 60% of consumers have used AI to help them shop.² In the B2B world, generative AI has overtaken Google as the starting point for a quarter of all buyers researching vendors.³

Trust hasn’t kept pace. A major 2025 international study spanning 47 countries found that only 46% of respondents globally are willing to trust AI.⁴ In the United States, the United Kingdom, and Germany, over 40% of the population explicitly resists increased AI adoption.⁵ American respondents are more than twice as likely to say they reject the growing use of AI than they are to embrace it.⁵

That gap produces a predictable behavior: customers use AI for speed and synthesis, then seek out human networks for validation. Eighty-five percent of users cross-check AI answers elsewhere rather than accepting them at face value.⁶ Among B2B buyers who encounter AI-generated summaries, 90% still click through to review sites and citations to verify the claims.⁷

This shows up in both consumer and B2B contexts. AI informs. People decide.

The Customer's Decision Ecosystem

Customers don’t make decisions alone. They draw on an ecosystem of sources, each playing a distinct role in gathering information, forming judgments, and ultimately committing.

That ecosystem has always included peers, colleagues, experts, and vendors. What’s changed is that AI now occupies a role too. Not replacing the others, but handling a specific kind of work.

AI: The Research Assistant

AI has earned a role in the ecosystem for a specific kind of task: fast, functional, information-dense questions where synthesis matters more than nuance.

Seventy-seven percent of consumers say AI helps them cut through information overload.⁸ Sixty percent feel it gives clearer answers than traditional search.⁶ When customers need to compare specifications, summarize options, or get up to speed quickly, they increasingly turn to AI first.

In practice, AI acts like a research assistant. Comprehensive, fast, and tireless. It answers the question: What are my options and how do they compare?

What it cannot (fully) answer is the harder question: Which option is actually right for me?

Peers: Still The Trusted Advisors

For that, customers turn to people.

Fifty-two percent of B2B buyers consult peers first when making purchase decisions. “Someone like me” remains the most trusted source of information.⁷ Edelman’s research found that peers are trusted at twice the rate of executives or officials, even when the topic is AI itself.⁵

The role peers play is trusted advisor. Not comprehensive, but credible. A peer can tell you what the implementation was actually like. Whether the support team is responsive. Whether the product works the way the demo suggested. These are lived experiences no summary can replace.

Fifty-four percent of B2B buyers have conversations with current customers before purchasing, often without the vendor’s knowledge.⁷ They are not looking for more information. They are looking for credible recommendations from “people like me”.

Communities: The Ongoing Context

Beyond individual peers, customers increasingly rely on communities. Private Slack groups, Discord servers, industry forums, subreddits. These spaces accumulate knowledge and context over time.

They work differently from one-off peer conversations. They provide persistent access to collective experience. A buyer can search years of discussions, see how opinions have evolved, and gauge sentiment over time. Communities act as long-term memory.

This is where quiet reference checking takes place. Buyers validate vendor claims in channels that vendors generally cannot see. The community becomes a trust layer that sits outside the vendor’s influence.

Again, the pattern: AI informs, humans decide.

AI as Accelerant, Not Replacement

Many executives are tempted to treat AI as a substitute for employees and to underinvest in, or divest of, the infrastructure (platforms and programs) that supports long-term relationships. Automate support, recommendations, and engagement. Reduce headcount while increasing efficiency.

This is a mistake.

AI earns trust for utility: speed, synthesis, cutting through noise. Human relationships earn trust for credibility: verification, judgment, and the confidence to act. These are not interchangeable.

Companies that use AI to attempt to replace relationship infrastructure risk being present for the research phase but absent for the decision phase. Customers will use AI tools to gather information, then turn to people to decide what to do about it.

The more strategic path is to use AI to make information accessible and reduce friction in discovery. Then invest more heavily in the communities and relationships that determine whether you move from shortlist to selection. And here’s the feedback loop: community investment has become an AI strategy. The companies with strong reputations in the spaces where real users gather will be the companies AI learns to recommend.

Implications for Digital Leaders

Your customers already participate in ecosystems of relationships, whether you realize it or not. They trust AI for some questions and peers for others. They discover communities that serve their needs better than mass platforms ever could. Leaders who understand this know where influence actually comes from. Those who don't will continue wasting money on channels that no longer influence decisions.

The question is whether you will build your strategy around how customers actually behave, or keep broadcasting into an ecosystem that is actively reorganizing, and in some cases, routing around you.

So, where do you start, and what should you do? My advice:

• Map your ecosystems. Where do your customers actually turn when making decisions? Which sources do they trust? Where are you present, and where are you absent? Most companies can draw their org chart and their tech stack. Almost none can draw the ecosystem of relationships that actually determines how their customers decide.

• Differentiate opportunities by utility or credibility: Know what you're competing on. Compete on utility where AI can help you: speed, synthesis, accessibility. Compete on credibility where only relationships can deliver: verification, judgment, and the confidence to act.

• Invest in the ecosystem. Seek out and participate in online and IRL spaces where your customers already gather. In addition to external opportunities, most B2B companies can host spaces, groups, and events to gather customers, prospects, partners, and employees. View community building as a capability. View your community ecosystem as a strategic asset. That's something competitors cannot replicate.

• Measure ecosystem reach, position, and value created. The old metrics measured broadcast: impressions, followers, and share of voice. The new metrics ask harder questions: Are you present where decisions happen? Are you consulted early or late? And are you contributing value to those spaces, or just extracting attention from them?

Treating customers as a passive audience should have ended when the Internet made conversation possible. Most companies never fully absorbed that lesson. They treated social media as another broadcast channel and optimized for attention instead of relationships. But customers were always participants in a system of relationships. The Internet made those relationships possible at a global scale, and AI has made them more visible and more complex, but neither has changed the fundamental reality: trust moves through relationships, and relationships are developed and nurtured by people.

The organizations that understand this will build infrastructure for long-term relationships. The ones that do not will keep making short-term investments for attention, wondering why the conversions never come.

AI informs. People decide.

Sources & further reading:
1. Search Engine Land, “37% of consumers start searches with AI instead of Google” (2025); Menlo Ventures, “2025: The State of Consumer AI”
2. University of Virginia Darden School of Business / Bloomreach consumer survey (2025)
3. Demand Gen Report, “GenAI Overtakes Search for a Quarter of B2B Buyers” (2025)
4. KPMG / Melbourne Business School, Global AI Trust Study (2025)
5. Edelman Trust Barometer Flash Poll: Trust and Artificial Intelligence at a Crossroads (2025)
6. Search Engine Land / Attest Consumer Adoption of AI Report (2025)
7. TrustRadius, “Bridging the Trust Gap: B2B Tech Buying in the Age of AI” (2025)
8. Darden Report / UVA Darden School of Business (2025)

Every year, you spend more to reach and reconnect with the same people.

Customer acquisition costs have increased 60% over the past five years (Genesys). Google's cost-per-click jumped 10% in 2024 alone, five times the increase from the year before (Wordstream). Your organic reach continues its long decline. Your engagement rates are falling. You're running faster to stay in place.

This is the treadmill of mass-market acquisition: every impression is a transaction that resets to zero. You pay for reach, you get a momentary glimpse, you start over tomorrow. The platform takes its cut. The algorithm changes. Your “audience” evaporates.

Meanwhile, something else is happening in the margins of your business. In customer advisory boards. In user groups and partner networks. In the Slack channels, WhatsApp groups, and private communities where your most engaged customers actually spend their time. Relationships are quietly compounding. People who trust you are introducing you to people who don’t. Yet.

One of these is a treadmill. The other is an investment.

The Social Depreciation Problem

The economics of mass social acquisition have fundamentally broken.

In my previous piece, The Last Days of Social Media, I documented the engagement collapse across major platforms: Facebook’s 0.06% engagement rate despite three billion users, Twitter/X’s decline to 0.029%, and Instagram’s 28% year-over-year drop. In mass social the engagement problem is really an economics problem. You’re paying more for less, and the less keeps getting smaller.

Consider what’s changed in the past five years. Customer acquisition costs have risen across nearly every industry. Digital advertising costs continue climbing as competition intensifies and privacy regulations limit targeting precision. The platforms that once offered organic reach have systematically eliminated it, converting what was once free distribution into paid media with declining returns.

This creates a structural disadvantage. Every dollar you spend on mass acquisition buys diminishing results. Every “audience” you build exists at the pleasure of the platform, subject to algorithmic changes you can’t predict or control. Every campaign starts from zero because impressions don’t compound. They decay.

The mass social economy runs on extraction. Your attention is monetized. Your engagement is harvested. Your relationships are converted into targeting data, which is then sold back to you at higher prices.

The Compounding Alternative

Relationship equity works differently.

When someone trusts you because of positive, consistent interactions over time, that trust becomes an appreciating asset. It reduces friction in every subsequent interaction. It increases willingness to pay. It generates referrals that bring in people who arrive already predisposed to trust you.

The research here is remarkably consistent.

Customers acquired through relationships and referrals are worth significantly more than those acquired through paid channels. A landmark study by Wharton researchers tracking 10,000 bank customers over six years found that referred customers remained more valuable throughout the relationship, with approximately 16% higher lifetime value than comparable non-referred customers. They were more profitable and more loyal, and the difference persisted over time.

Then there is the classic finding from Frederick Reichheld’s research at Bain in the 1990’s that a 5% improvement in customer retention can drive profit increases of 25% to 95%. Not because retention is cheaper than acquisition (though it is), but because retained relationships compound in ways that new acquisitions cannot.

Here’s what compounding looks like in practice:

• Depth. The longer you know someone, the more context you share. You understand their situation, challenges, and needs. They understand your capabilities. Every interaction builds on previous interactions rather than starting fresh. Cost-to-serve drops as customer knowledge and shared trust accumulate.

• Breadth. Trusted relationships generate new relationships. People recommend businesses to people they know and trust. Research consistently shows that referred customers convert faster, cost less to acquire, and demonstrate higher loyalty. Referral-based acquisition costs a fraction of paid channels, often $150 compared to $800 or more for paid search in B2B contexts.

• Resilience. Relationships survive competitive pressure. Eight in ten consumers say they’ll stick with a brand they trust even when another brand becomes “hot and trendy.” They’ll pay more. They’ll forgive mistakes. They’ll advocate on your behalf. This is the opposite of the algorithmic economy, where every interaction is a fresh evaluation and switching costs approach zero.

Relationship Equity Compounds Across the Customer Lifecycle

Executives often (and unfortunately) file "relationship strategy" under customer success, somewhere between onboarding and churn prevention. That's a fraction of the value. Relationship equity transforms how you acquire customers, engage them, expand accounts, and ultimately, grow your market share.

• Acquisition: Your best customers become your most valuable growth channel. When referral-based CAC runs five to six times lower than paid acquisition, the math reshapes everything. You’re not just saving money. You’re acquiring customers who arrive with built-in trust, shorter sales cycles, and higher lifetime value. The community ecosystem generates growth that paid media cannot replicate.

• Engagement: Connected customers don’t just consume. They contribute. They help each other succeed. They generate insights you couldn’t buy through research. They stress-test your assumptions and surface problems before they become crises. McKinsey research shows that 84% of Gen Z consumers trust product reviews from niche online communities more than corporate advertising. The ecosystem creates value that the organization alone cannot.

• Expansion. Trust is the foundation of growth within accounts. Upselling and cross-selling happen because of relationships. When someone trusts you, they tell their coworkers and peers. Purchase decision makers are willing to extend that trust to adjacent offerings. When they don’t trust you, every new conversation starts from scratch.

• Advocacy. Connected customers recruit. Three-quarters of consumers say they'll recommend a brand they trust when asked. Every recommendation has the potential to generate new customer relationships. Paid reach resets to zero each morning. Advocacy accumulates.

Executives often frame acquisition and retention as competing budget lines. Relationship equity makes the question irrelevant. Your ecosystem acquires and retains simultaneously.

The AI Inflection

Two converging forces make relationship equity more valuable than ever.

First, content is becoming commoditized. Every organization can now generate infinite content. The marginal cost of producing another blog post, another social update, another email sequence is nearly zero. What was once a competitive advantage, the ability to consistently produce and distribute content, has become table stakes.

When everyone can produce infinite content, attention fragments across endless options. The scarce resource becomes the trusted voice that helps you filter what matters.

This is where I want to introduce a concept I’m calling relational trust. The idea goes beyond someone trusting you in this moment. It’s that they trust you because of accumulated experience over time: shared context, demonstrated reliability, history that cannot be manufactured or accelerated.

AI can produce content. It cannot produce relational trust. An algorithm can optimize for engagement. It cannot replicate the relationships you’ve built through years of consistent interaction. A competitor can copy your messaging. They cannot copy your ecosystem.

Second, first-party relationships are becoming first-party data. Third-party cookies are dying. Privacy regulations are tightening. The organizations with direct, trusted relationships with their customers hold an asset that cannot be replicated by competitors or regulated away.

The value of relationship infrastructure extends beyond engagement to intelligence. When you have direct relationships with your ecosystem, you understand your market in ways that surveillance-based analytics can never match. You hear concerns before they become complaints, spot trends before they hit the broader market, and know what your customers actually think because they tell you.

Your “network of relationships” is a strategic asset that appreciates over time. And unlike reach or impressions, it cannot be taken away by an algorithm update or a platform policy change.

From Audiences to Ecosystems

The mindset shift I’m advocating is in how senior leaders think about growth.

Mass social media trained us to think about digital networks in terms of audiences: passive recipients of one-to-many communication, counted by reach and impressions. The relationship-first model brings us back to the more meaningful opportunity: interconnected networks where communities form and value flows in multiple directions.

Audiences are rented. Ecosystems are hosted (and owned). When you build an audience on a social platform, you don’t own the relationship. The platform does. They can change the rules, limit your reach, or shut you down entirely. And keep in mind, these are often customer relationships you already have access to. When you build an ecosystem of direct relationships, that infrastructure belongs to you.

Audiences decay. Ecosystems compound. Every social media impression starts fresh. Every relationship builds on what came before. The difference is between a transaction and an investment.

Audiences are measured by reach. Ecosystems are measured by connection. The metrics that matter aren’t how many people saw your content. They’re the density of relationships within your network, the velocity of referrals, the depth of trust that enables everything else.

This doesn’t mean abandoning all mass-market activity. It means recognizing that mass-media activity should feed ecosystem development. Use reach to attract. Use relationships to retain, grow, and generate the next wave of attraction.

The Compound Interest of Connection

The organizations that thrive in the next decade won’t be the ones with the biggest audiences. They’ll be the ones with the most robust relationship networks: ecosystems where relational trust compounds, where customers become advocates, where every connection has the potential to generate the next one.

You can keep renting attention on platforms that take more and give less every year. Or you can start building the relationship infrastructure that appreciates over time.

Every customer interaction is either a transaction or an investment in a more connected relationship. One resets to zero. The other compounds.

Early in the pandemic, I started a weekly call for community leaders. Monday mornings. No agenda. Just friends and acquaintances in the industry trying to figure out how to keep ourselves sane and keep our communities connected while everything in the real world shut down.

What became apparent in the early days of the pandemic lockdown was how thin our existing tools were. Forums felt static and one-dimensional. Social feeds felt noisy and hollow. People weren’t looking for more content. They were looking for presence. Real conversation. A sense that someone else was actually there, in that shared moment in time.

The pandemic didn’t so much break social media as deeply expose its flaws.

The original promise of social media was a big tent where everyone was welcome.

Instead, the mass social platforms became very good at capturing attention and very bad at supporting human connection. The reality of mass social media is a marketplace of attention, behavior, and data, and the clear winners are the platform providers that harvest and monetize the marketplace.

Mass social media had trained us to confuse activity with connection. Scrolling replaced conversations and connection. What we experienced was a facsimile of real life, as algorithms shaped our experiences. In effect, through social media platforms, we were each in a community of one.

Cut to December 2025: How is all this working out?

While most people with an Internet connection also use social media, the primary activity is consumption. Participation and engagement are evaporating.

• Facebook maintains 3.07 billion monthly active users, (CNBC) yet average page engagement sits at a dismal 0.06%. Instagram’s 2-3 billion users (Electro IQ) generate only 0.36-0.50% engagement—a 28% year-over-year decline. (Socialinsider) Twitter/X presents the starkest case: engagement rates have collapsed to 0.029% across industries (SocialSellinator), with users posting 38% less frequently than in 2023 (TECHSABADO) and daily active users falling from 250 million to 157 million under Musk’s ownership. Epic Owl

• TikTok remains the outlier with 2.5% engagement rates (Electro IQ) and 58-95 minutes of daily usage—yet even this platform shows a decline from its 5.14% peak. (Social Media Today) LinkedIn has bucked the trend entirely, with engagement climbing from 6% to 8.01% (Buffer) as professionals migrate serious discourse away from deteriorating platforms. But even LinkedIn isn’t immune from controversy, as recent changes have signifigantly negatively impacted organic reach for most contributors.

• Time spent tells a more nuanced story than user counts suggest. Global social media consumption has dropped approximately 10% since 2022, now averaging 2 hours 20 minutes daily. (RTÉ) More critically, the nature of that time has shifted: 68-78% of social media activity is now passive scrolling rather than active engagement. (Powertechjournal) Users maintain accounts but have stopped posting—a phenomenon researchers call “lurking” that represents the worst outcome for platforms dependent on user-generated content.

Gen AI hasn’t helped the quality of content and engagement on mass social platforms, as synthetic profiles shovel AI-generated text, audio, and video into feeds.

You can see the fallout in how people talk about social media now. Breaks are framed as recovery. Scrolling is described as a habit (doom scrolling), not a pleasure. Posting feels risky, while silence feels safer.

The open and authentic contributions aren’t slowing down because people are “disengaged”; it is because they are exhausted from being part of the social media machine.

The time is right to rethink community at the human scale

The collapse of mass social engagement is a correction. For two decades, we operated under the assumption that bigger was better, that scale was the ultimate metric of success, that the global village could contain meaningful connections among billions. That assumption was always suspect. Now we have the data to prove it.

Here’s what the numbers tell us: while mass platforms struggle to crack 1% engagement, private communities routinely achieve 40% monthly engagement rates. Discord’s own research shows that 90% of private servers have fewer than 15 members, and servers with 50-100 members achieve 78% engagement. The platforms built for mass scale are being used for intimate connections. Users are telling us something.

The research on optimal community size has been remarkably consistent for decades. Robin Dunbar’s work on cognitive limits suggests we can maintain meaningful relationships with roughly 150 people, and that’s the outer edge. Our circle of close relationships tops out around 15. This is a feature of human cognition to embrace. Communities can build around this limit rather than fighting it.

What we’re witnessing is a migration already in progress. Eighty-four percent of online sharing now happens through “dark social” channels: private messages, group chats, curated spaces invisible to the algorithmic eye. People haven’t stopped connecting. They’ve stopped connecting in public, on platforms that monetize their attention and weaponize their data.

From audiences to relationships: The strategic pivot

In 2018, I developed a framework for understanding how community strategy evolves within organizations. Most companies start with what I called “function-centered strategy,” where community serves as a support deflection mechanism, measured in tickets avoided. Some progress to “experience-centered strategy,” where community supports discrete customer touchpoints, generating leads and user-generated content. A few reach Essential Community, a “relationship-centered strategy,” where community becomes the medium for nurturing connections throughout the customer lifecycle.

The next shift, the one that’s now both possible and necessary, is toward what I called “transformational strategy.” This is where the network of relationships catalyzes a transformation of the business itself. Community becomes the dominant expression of brand, core to the product offering, and expected by customers.

Seven years later, the case for this transformation has only strengthened. Consider what’s changed:

• First-party relationships have become first-party data. Third-party cookies are dying, and privacy regulations are tightening. The organizations with direct, trusted relationships with their customers hold an increasingly valuable asset. A curated community is both an engagement channel and a proprietary intelligence layer.

• Customers seek more than products. They seek purpose and belonging. The research is consistent: people join communities for information but stay for identity. They want to be part of something, to contribute to something, to matter to someone. Mass platforms weren’t designed to deliver this.

• AI is commoditizing content. Every organization can now generate infinite content. What they can’t generate is trust, relationship, context, and history. The value is shifting from what you publish to who you know and how well you know them.

Successful communities operate on the economics of relationships, not attention. The value compounds over time rather than decaying with each algorithmic update. Members contribute because they're invested, not because they're being manipulated. Trust accumulates. Context deepens. The longer someone participates, the more valuable their participation becomes, to them and to the community.

Building the case for investment

If you’re reading this and thinking about how to make this case within your organization, here’s the argument in its simplest form:

The public platforms where you’ve been investing are declining. Engagement metrics are collapsing, and so is fundamental value. You’re paying more to reach fewer people who trust you less. The algorithmic advantage has shifted entirely to the platform, not the brand.

Meanwhile, owned community infrastructure, where you control the experience, own the relationships, and compound value over time, has never been more accessible or more necessary.

The question is whether you’ll make this shift proactively or reactively, whether you’ll build your community before or after your competitors build theirs.

What this looks like in practice

The transition from mass social investment to curated community investment requires shifting the strategic center of gravity.

• Start with purpose. Before you choose technology, get clear on why people would want to be in a relationship with you and each other. What transformation are you enabling? What is the larger whole that you are encouraging members to belong to? What problems are you solving together that can’t be solved alone?

• Think relationships. Stop counting followers and start understanding the actual network of connections among your customers, partners, and stakeholders. Who knows whom? Where does value flow? Where is there latent relationship potential you’re not activating?

• Design for the human scale. Use Dunbar’s Number as a design principle. Create structures that enable intimacy within larger communities: cohorts, circles, pods, chapters. Let people find their 15 within your 1,500.

• Invest in qualification over growth. The value of a community is determined by who’s in it, not how many. A vetted, qualified community of 500 engaged members will outperform a mass audience of 50,000 passive followers every time.

• Relational measurements over time. Move beyond reach and impressions to relationship metrics: trust, reciprocity, contribution quality, network density, and member lifetime value. The metrics of mass social don’t transfer to relationship infrastructure.

The last days and the first days

We’re living through a transition that will seem obvious in retrospect. Mass social media functioned brilliantly as an attention and data extraction mechanism. It has failed as a model for human connection.

What we needed, what we’ve always needed, is what we had in those Monday morning calls during the pandemic. Presence. Real conversation. The sense that someone else was actually there.

The last days of mass social media will also (hopefully) be the first days of something more human: smaller spaces, slower conversations, deeper trust, and a conscious choice to build digital infrastructure that serves human connection rather than exploiting it.

The tools exist. The economic case is clear. The migration is already happening. The only question is whether you’ll be building the communities people are moving toward, or watching from the platforms they’re leaving behind.

And lastly, know that we still have power as individuals and leaders.

As Howard Rheingold originally said about early communities, like The WELL:
”What it is, is up to us.”

While executives write checks for “community initiatives,” most organizations are still running the same tired playbook: pick a platform, hire a community manager, count the likes, hope for magic. The defaults have become so ingrained that we’ve stopped questioning whether they actually work.

They don’t.

I’ve spent 25 years building communities for companies like Autodesk, Dell, IBM, and Fast Company. I’ve watched the rise and decline of social platforms. I’ve seen millions invested in community theater that delivers no strategic value. And I’ve seen what happens when you reject the defaults and actually design community as a strategic system.

The difference is profound.

The Default Trap

Here’s what default thinking has historically looked like:

Platform-first planning. Teams chose an Enterprise “default” platform like Khoros, or before asking what relationships they’re trying to build. The tool becomes the strategy.

Activity metrics over value creation. They count posts and members while ignoring whether any of it connects to business outcomes.

Community as a cost center. It is sequestered in support, disconnected from other customer-facing functions like sales, marketing, and customer success.

Short-term thinking over long-term relationships. Quarter-focused initiatives for relationships that last years.

These defaults exist because they’re easy. They’re what vendors sell. They’re what conferences and gurus teach. They’re what your competitors are doing.

They’re also why most community strategies fail.

Community as Strategic Architecture

Real community strategy starts with a completely different mindset: community is relationship-building infrastructure, not a marketing channel.

Think about your most valuable customer relationships. They don’t live in a single platform or program. They exist across an ecosystem of events, networks, advisory boards, user groups, and certification programs. Strong relationships create compound value over time. They reduce support costs. They accelerate product adoption. They generate referrals. They provide market intelligence.

But only if you design for it.

The Essential Community Strategy Framework treats community as a strategic architecture that requires the same rigor as product development or go-to-market strategy. It’s built on seven stages that move from insight to action:

• Commitment – Not just budget, but organizational alignment. The right sponsors, stakeholders, and structure to do meaningful work.

• Research – Deep understanding of member needs, ecosystem dynamics, and value exchange. Not surveys, but actual insight into how relationships create value.

• Definition – Clear purpose and vision that aligns member value with business value. No generic mission statements.

• Exploration – Opening the aperture to see the full range of possibilities. Looking beyond the platform to the ecosystem.

• Selection – Choosing programs based on strategic value, not tactical convenience. Building a portfolio, not a feature list.

• Implementation – Operating model that enables distributed execution with a centralized strategy. A Community Center and Programs Office structure that scales.

• Evaluation – Qualitative signals at the macro level, quantitative metrics at the program level. Learning loops, not just dashboards.

Why a Design Approach is Essential

Three forces are converging that make this the moment to rethink community strategy:

• Social platforms are collapsing. The spaces where we built communities are fragmenting. Twitter/X is unrecognizable. Facebook groups are ghost towns. LinkedIn is a content mill. Your members are looking for something better.

• Ecosystem thinking is ascending. The smartest companies recognize that value creation happens across networks, not within silos. Community is how you activate and orchestrate ecosystem relationships.

• AI makes community more valuable, not less. As content becomes commodified, authentic human relationships become the differentiator. Community provides the context, trust, and feedback loops that AI cannot. And community will become one of the most valuable sources of primary data for companies.

The companies that recognize this shift will build sustainable competitive advantage through their community ecosystems. The ones that don’t will keep funding failed initiatives and wondering why their NPS scores keep dropping.

From Default to Design

I developed this framework through hundreds of strategy engagements, watching what actually works versus what we think should work. It’s been pressure-tested with enterprise software companies, membership organizations, and platform businesses.

The organizations that succeed with it share three characteristics:

1. They reject the premise that community is a nice-to-have. They see it as strategic architecture for customer relationships.

2. They commit to the full cycle, not just the fun parts. Research isn’t optional. Governance isn’t an afterthought. Measurement isn’t just vanity metrics.

3. They build for the long term. Community strategy is a practice, not a project. It evolves as your business, market, and ecosystem evolve.

The Choice

You can keep running on defaults. Keep funding platforms that don’t deliver value. Keep wondering why your community initiatives fail to move the needle.

Or you can design something better.

The Essential Community Strategy Framework gives you and your team a structured approach to move from default thinking to intentional design. To build community ecosystems that create compound value over time. To transform community from cost center to strategic asset.

The framework deck is available here. It includes templates, worksheets, and tools to guide your team through each stage.

But more importantly, it includes a different way of thinking about community. One that treats relationships as infrastructure worth designing, not accidents worth hoping for.

Your customers are looking for meaningful connections. Your business needs sustainable differentiation.

It’s time to stop accepting defaults and start designing for what’s actually possible.

Bill Johnston is Founder & Community Architect at Structure3C. Over the past 25 years, he has designed community strategies for Autodesk, Dell, IBM, Fast Company, and dozens of other organizations. He works with leadership teams to build community as strategic infrastructure, not tactical add-ons.

I’ve been studying the role AI can play in community and customer experiences for several years. In our first round of research in 2018, AI was still a very abstract concept that most people viewed as an analytics tool. More recently, LLMs like ChatGPT have helped many of us go from needing to be best friends with a data scientist to actively learning by exploring new tools, experimenting with prompts, and tinkering with agent workbenches.

The Lab at my September event was intended to move participants from undirected experimentation (prompt play) toward meaningful experiments, prototypes, and pilots. I wanted to help teams explore where AI could truly improve the customer or member experience and give them space to test those ideas and refine them quickly.

How the Lab Worked

We began with an intentionally simple journey map of a community member lifecycle: Onboarding, Participation, Contribution, and Advocacy.

Each team chose one stage and then brainstormed specific pain points or opportunities that exist for their inidividal communities. For example: maybe onboarding felt impersonal. Maybe participation had dropped off. Whatever the issue, the question was, How might AI help us design a better experience?

To get people thinking differently, I invited them to picture AI as a specialist or a small team of specialists that they could collaborate with. I gave examples like a researcher who could synthesize large amounts of data, a designer who could draft concepts on demand, or a coach who could personalize guidance for each member. This framing helped everyone see AI as a creative teammate rather than a technical feature.

Teams used ChatGPT, Gemini, and Claude to explore ideas, develop content, and build quick prototypes.

Each group produced two outputs:

1. A concept poster that included a problem statement, a short headline describing the value of their idea, and a rough sketch of how AI fit into the experience.

2. A prototype artifact created with AI, which included a personalized onboarding messaging framework, a recognition workflow, or a dashboard concept.

In less than ninety minutes, the room was filled with flipboard sketches, post-it notes, and framework drafts. The animated conversations from each group were often complemented by one or more ChatGPT voice chats happening simultaneously.

What We Learned

Reflecting back on the experience, several things stand out:

• Each group began moving quickly because they all had a clear frame. By anchoring the work in a familiar journey and giving AI a practical identity (a capable teammate with defined skills ) people found it easier to generate ideas that felt real.

• Participants came in with very different levels of AI fluency and different tool preferences. Some were already using these tools daily. Others were less frequent users. That gap mattered. In the future, I’ll send a short “warm-up” before the session so everyone arrives with a shared baseline.

• The third lesson was about time. I wanted to use the relatively short session window (90 minutes) to provide a sense of urgency and to act as a design constraint. While teams created strong concepts, we ran out of time to connect their ideas back to measurable outcomes. The creative work was there, but understandably, the AI artifacts were just initial drafts.

• Perhaps the biggest takeaway was how naturally people collaborated with AI once they stopped treating it as a machine to be prompted and started treating it as part of the team.

Time To Explore

We are still early in the process of figuring out how AI will impact customer and community experiences. What matters now is creating the time and space to learn. Moving beyond “prompt play” involves working on real-world challenges, giving AI tools a role, and bringing them into the research and design process.

As we get better interfaces and experiences from the AI tools, it will also become easier to involve business stakeholders and customers/community members as direct participants in the process.

Next Steps

The Lab was a reminder that progress starts with curiosity. When we ask better questions and give ourselves time to explore, we create the conditions for answers to materialize.

A Simple Exercise for Teams

Here’s a simplified version of the AI Lab session you can try with your own team. It works well in a ninety-minute meeting or workshop.

Step 1. Frame the Challenge

Pick one stage of your customer or member journey.
Define one pain point or opportunity, like “members drop off after sign-up.”

Step 2. Imagine AI as a Team of Specialists

Ask your group: If we had a team of AI specialists supporting this work, what kinds of expertise would we want or need?
Think about researchers, analysts, writers, designers, or coaches. Describe how these specialists might work alongside your human team.

Step 3. Co-Design with AI

Use AI tools in progressive loops to generate ideas, refine ideas, and develop draft content, artifacts, or simple prototypes.
Remember: Keep the focus on customer impact, not on tool features.

Step 4. Capture and Share

Document your concept in two ways:

1. Concept Summary – a short problem statement and headline (“Personalized onboarding reduces drop-off by 25%”).

2. Prototype Artifact – a tangible output created with AI, such as a draft messaging framework, experience flow, or dashboard.

Step 5. Reflect and Plan Next Steps

Review with the group:

• What surprised us?

• How might we test this in the real world?

• What would we need to learn before testing this with customers?

What I saw in the room at our September event helped me see a range of new possibilities for working with AI in a group setting, in real-time, and giving AI an active role as a collaborator. I’m excited about the possibilities.

If you are interested in exploring this with a group of smart and passionate community and customer experience professionals, sign up below. We hold monthly online events, and our next in-person event will be in March 2026.

Sign up here to be considered for future Structure.Community events.