The Case for AI Practice Groups

We’re almost halfway through 2026. Let that sink in.

January gave way to spring and then to summer in what felt like two weeks. In the same two blinks, it will be October heading into the holidays.

By then, more likely than not, your board will be asking how to measure AI ROI from the initiatives they approved this year. The latest reports show that top-down, technology-driven programs are not delivering it. Most AI projects fail not because the models are weak, but because organizations deploy AI without redesigning how work actually gets done. AI pilot programs failing has become the norm, not the exception.

At the same time, a different problem is building from below. Last month’s booing of commencement speakers talking about how AI is changing work is emblematic of real misgivings among rank-and-file employees about whether the benefits outweigh the costs. There is growing resentment of executives pushing top-down AI programs that feel imposed rather than earned. A 2026 survey from Writer found that 29% of employees, and 44% of Gen Z, admit to actively undercutting their company’s AI strategy: slow-walking it, refusing to use the tools, even feeding it bad inputs to make the AI look worse than it is. Alan Berrey described this dynamic in detail. When fear stays unspoken, adoption stalls and pilots die without anyone naming the cause.

You’re getting squeezed from both ends. The top and the bottom.

The good news is that the solution to both problems may already exist inside your company.

What Are AI Practice Groups?

The idea isn’t new. Etienne Wenger and William Snyder were writing about communities of practice in Harvard Business Review back in 2000, describing the informal networks that naturally form when people share a craft and care about getting better at it together. What’s new is what they’re forming around.

AI Practice Groups are the practitioners doing the real work of adoption: small groups of employees who are all-in on AI, swapping stories, comparing tools, and finding ways to use new capabilities in actual workflows. Sometimes management sponsors them. Often they form on their own, as early-adopter practitioners find each other and trade failed experiments and hard-won discoveries.

These aren’t token-maxing engineers. They’re regular employees on the front lines comparing vibe-coded apps, arguing about frontier models, debating when to switch from text to voice mode, and figuring out how to use tools like Cursor and Cowork in real work, not just to plan vacations or shop for running shoes.

Seampoint’s model organizes serious AI work into three groups. The AI Council sets strategy and owns metrics, often an executive steering committee. And the AI Guild validates tools and sets the standards, often the IT department . But the AI Practice Group is where value actually gets created, where a licensed tool becomes an everyday habit in your organization. Nearly every company gets the sequence wrong. They stand up the Council first, then the Guild, and only eventually reach the practitioners. That top-down AI adoption strategy for business feels orderly. It also produces nothing if the herd never actually grazes the grass.

Here’s a bold idea: start with an AI Practice Group. Get your AI boosters together. You can find them at the water cooler or through an informal company survey (see below). Then watch and listen. Build the Council and Guild on what the practitioners reveal. The handful of companies actually capturing returns aren’t the ones with the biggest central AI labs. They’re the ones that pushed AI out to empowered teams closest to the work line managers and practitioners, not strategy decks in a conference room.

Why AI Practice Groups Solve Both Problems

Three reasons:

1. They’re already invested. These employees are spending their own time above and beyond their workload. They’re vested in making AI work, not waiting for permission.
2. They produce bottom-up ROI. They represent the kind of applications to real company workflows that can actually return value, unlike top-down initiatives that stall and get undermined. This is the difference between panning for gold and mining it.
3. They have peer credibility. If anyone is going to persuade the large share of your workforce that is wary of AI, to say nothing of the hard core who are flatly against it, it will be their peers, not their bosses. The wariness is real and well documented: Pew found that about half of U.S. adults are more concerned than excited about AI’s growing role in daily life, and only about one in ten are more excited than concerned. And the research on what actually shifts that wariness is clear. An HBR piece this March by Baym, Dillon, and Jaffe found that peer influence, not top-down mandates, is what moves adoption at scale. A well-facilitated AI Practice Group puts uncertain employees shoulder to shoulder with co-workers facing the same tasks, until the threat in their imagination shrinks to the size of the actual problem in front of them. Done wrong, the same group can intimidate the uncertain majority with demos and jargon. Done right, with a sentinel steering toward shared problems rather than individual performances, it calms the herd.

Why Informal AI Practice Groups Can Make Things Worse

That said, it’s important to realize that this same grass-roots energy, without structure and facilitation, can actually deepen the problems you’re trying to solve. The risks are not primarily about data leaks or bad prompts, though those matter. They’re about what an unstructured group fails to build.

1. Judgment debt. The company’s biggest exposure is not a single bad output. It is the accumulating gap between what employees confidently delegate to AI and what they actually know how to verify, escalate, and own. An informal AI Practice Group swaps tool tips but never pays down the debt. Members leave knowing a new trick, not a repeatable judgment about when to trust AI, when to check it, and when to stop. A properly run AI Practice Group is where that debt gets worked off, one real decision at a time.
2. Wrong incentives. Loosely organized groups tend to become trick-of-the-week sessions. Members show up to impress each other with the latest demo or video, not to work through the failures, edge cases, and verification steps that actually matter. Enthusiasm substitutes for progress.
3. Hidden risk. Poorly facilitated groups celebrate wins and skip the post-mortems. Nobody brings the output that looked right but wasn’t, the workflow that broke quietly, or the prompt that sent sensitive data somewhere it shouldn’t have. The group feels productive while the real exposure stays invisible.
4. No telemetry. Management gets zero usable feedback. Leadership hears that “people are excited about AI” while remaining blind to what is landing, what is stuck, and how employees actually feel about the work. Without that signal from the grass, the AI Council is guessing.
5. Amplified resistance. An informal group of power users can widen the gap between believers and skeptics rather than close it. Employees who were already unsure watch colleagues demo capabilities they cannot match and conclude AI is for other people, or that the company is leaving them behind. Fear actually increases.
6. No compliance record. A group that meets without structure, documentation, or a shared method leaves no defensible record of who decided what, on what basis, and against which framework. When something goes wrong, or when a regulator, insurer, or board asks what your organization actually did, there is nothing to show. That gap between what your people are doing with AI and what your policies protect is widening fast.

Every AI Practice Group needs a basic AI governance checklist and a real AI implementation framework, not just slides, but a bona fide operating rhythm with a facilitator (or “sentinel”), shared modules, and records that track the work being done. Without that, you are most likely amplifying the problem and calling it progress.

How to Invest in Your AI Practice Groups

Level up what you already have. Give your groups four things:

1. Identify them. Step back from day-to-day work and survey the team. Ask who is using AI tools and what for. Talk to those people. Learn which tools they use, who owns the accounts, and whether enterprise controls and zero data-retention agreements (ZDRs) are in place. Treat this as a lightweight AI readiness checklist not a bureaucratic assessment, but an honest inventory of what’s already happening in the shadows.
2. Formalize their meetings and give them a sentinel. Provide a forum, a regular cadence (weekly or bi-weekly), topics of discussion, and follow-up. Assign a facilitator who can sustain the rhythm when enthusiastic volunteers burn out, ideally someone with experience guiding practitioners through this terrain who can carry proven modules and session design so the group compounds rather than fizzling. The sentinel’s job is not to run a presentation every week. It is to identify AI workflow failures and wins by stimulating conversation, steering the group away from trick-of-the-week demos, and building a record of judgment through real case studies from your company. Start examining AI-assisted workflows for consequence of error, verification cost, and necessary authorizations.
3. Connect them to executive infrastructure. The AI Practice Group’s learnings need a path to the wider organization and a record that survives the meeting. That means executive sponsorship, structured telemetry back to leadership, and documentation that creates a compliance record (i.e. who decided what, on what basis, and what the group learned). This is where the AI Council earns its keep. Not by dictating the rules, but by listening to what practitioners are actually demonstrating.
4. Propagate the model. Once one AI Practice Group is working, replicate it across other departments using the same structure.

If at all possible, avoid subjugating this work to IT or trapping it in HR. The AI Practice Group’s value is peer credibility on the front lines. Buried inside a ticket queue or a training calendar, it loses its power.

---

The operational playbook is simpler than the problem feels. You don’t need a grand strategy or a bet you can’t reverse. You need one AI Practice Group in motion, a sentinel keeping it on course, and an AI adoption strategy for business that starts where value is actually created, at the front lines of your organization.

If you already have green shoots like an informal Slack channel, a brown-bag group, or a handful of believers, you’re further along than you think. Learn more about how Seampoint helps organizations assemble and equip AI Practice Groups.