The true test of adoption is whether the New Thing – the tool, process, or practice you’ve incorporated into your workflows – actually makes your team better – not just more informed. And articifial intelligence is no exception to this rule.

These kinds of rollouts often lean on courses, wikis, and launch emails to socialize how they work and what value they might bring to the team. Useful for context? Sure. But our data shows they’re not so effective at changing daily behavior. In our latest AI Collaboration Index, the two most common tactics – formal training courses (69%) and self-serve knowledge hubs (57%) – are among the least effective at improving the way people actually work. Meanwhile, teams that make time to experiment report a 21% bump in creativity, and they tend to win back hours as those experiments harden into established norms. And that pattern holds beyond AI: protected, hands-on trial and error beats passive consumption when the goal is real, useful adoption.

BUCK – the design company behind work for Apple, Google, and Airbnb – learned this firsthand. Describing how their creative team adopted generative AI across production – both behind the scenes and in shipped client work – Chief Creative Officer Kevin Walker explains, “If you’re just broadcasting, just saying something, it’s not as impactful as the act of doing.”

The broadcast trap

When leaders feel urgency, they often tend to double down on artifacts they can control: internal announcements, fancy decks, trainings and more trainings. And these mechanisms have a place – they establish guardrails, clear policies, and a shared language – but they’re rarely effective in creating new muscle. People adopt a practice when they try it, it works (even a little), and they see someone they trust doing it for real.

That last piece is the unlock. Teams are four times more likely to make time for their own experiments when a manager demos live. Not a keynote – an honest-to-goodness screen-share of a real workflow, narrated in real time. So the way out of the “broadcast trap” isn’t a better memo. It’s a smaller bet, shown at face value.

Make space for practice

BUCK found progress only arrived once they literally made room for it – time on the calendar, permission to be messy, and a clear expectation to share the result. “Without carving out space and time,” Kevin says, “you’re not gonna get anywhere.” Because once that space exists, experimentation starts feeling less like extracurricular chaos and more like a productivity engine. Even before you’ve nailed the routine, you leave with something useful – a concrete example of what changed, something small you can reuse, and a clearer idea of how to start faster next time.

Let’s call these recurring slots “practice blocks”: simple, repeating calendar holds where a small group takes one real workflow, tries one new move, then shows the work. Keep the scope small and the stakes low so the focus stays on learning, not performance.

To make each practice block pay off, slow down for a minute to capture and share:

  1. Post the proof. Share a quick “before vs. after” snapshot – say, two screenshots, or a 30-second Loom – in a place your team already checks (your main team channel or a shared doc). Add a sentence of context: what you tried and where.
  2. Save one artifact. Paste the exact prompt, checklist, or snippet into the same post, and add a one-line “how to use this” note. If it’s a template, drop it in a shared folder so others can find and reuse it.
  3. Write three lines. Under that post, add three bullets labeled Keep / Change / Kill that spell out what you’ll do next time – plain language, no jargon.
  4. Tag an owner + date. Name who to ping with questions and when the next practice block happens.
  5. Queue the next move. One sentence on what you’ll try in the following block to keep momentum going.

FOR EXAMPLE

  • Have AI draft a first-pass kickoff agenda with time boxes based on your project description. Tweak it and send it as the invite.
  • Paste a recurring task you do and ask AI for a one-click checklist to reduce mistakes. Try the checklist once and edit what didn’t help.
  • Ask AI to role-play a skeptical customer and chat through your new feature for two minutes. Borrow one stronger reply and add it to your response library.

The highest-leverage move for managers

Managers, if you only do one thing, do this: demo your own work live.

Choose a workflow you truly own and run it live for about 15 minutes, trying one new move. Talk through what you’re doing as you go: why this approach and what you’re looking for. Show the parts you still check by hand, and don’t tidy away the messy bits. Let the misfires and the recovery moves play out on screen.

That single ritual does three things:

  1. Signals permission – experimentation isn’t extra credit; it’s how we improve.
  2. Lowers the bar – no one has to be perfect to participate.
  3. Creates demand – a real five-minute gain makes others want to use the same shortcut.

You don’t need a giant rollout plan for this. You need a calendar invite with your name on it. If you like, enlist a couple of peer-credible doers to co-host early sessions, but keep the spotlight on manager modeling – that’s the accelerant.

Small experiments yield meaningful payoffs

Low-lift experiments are cheap to try and quick to judge. They create social proof (we saw it work), tangible assets (we can reuse it), and a short path to value (we can apply it tomorrow). That’s why they outperform “learn first, maybe try later” approaches – especially when the ground is moving. It also explains the numbers: make time to experiment and you get that 21% creativity lift; demo live as a manager and your team is four times more likely to make time to try things themselves.

Protect this time like any other critical work. As Kevin reminds us, “We need to make sure that we are really fighting so that people have room to transform – to adopt and adapt.” 

Avoiding the broadcast trap: A simple ritual for effective adoption