When leaders describe how their team uses AI, I often hear the same picture. A few people use it all the time and are building things with it. A few barely use it. And everyone else has tried a few things here and there. The leader wants the whole team using it, and their question is how to bring the rest along.
My answer is to show each person what AI can do with their own work, then give them time to go deep on one piece of it. Seeing AI handle their own work is what gets people to lean in, and it also explains why most of the team hasn't yet.
Why most people hold back
Ask people why they haven't experimented further with AI, and a lot of them put it down to not having the time or the energy. But the people I've seen get the most out of AI are usually the ones who were shown entirely new ways to work with it, on their own work, and then had room to experiment. Being told that AI can do all sorts of things doesn't help much, because it's too general to act on.
And the numbers point the same way. In a Gallup survey of US employees this year, among people in organizations that make AI tools available, 39% of those who don't use AI at all said they don't believe it can help with the work they do. The way past that belief is to show people AI working on their own work that they own and operate daily.
Show it on their own work
Take a finance team. What gets them interested is seeing the reports they compile at the end of every close, and the coordination with the board that follows, done faster and to a higher standard, with time left over for other work.
That's how I build the workshops I run for teams, in three steps.
- The first session gives everyone the same baseline, a shared language and a working sense of what today's AI tools can do.
- Between sessions, each person picks something in their own work they want to solve.
- In the next session, people bring those problems, and we work through some of them with AI in front of the whole group, so everyone sees how a colleague's problem gets solved.
Flare, a legal tech company, shows how quickly that can change someone's mind. I ran two workshops there, a week apart, for a group of about 15 people from across the business. Before the first, one team member told the group he was never getting rid of Excel. The team used Claude as its AI tool, and as I wrote in the case study, "Just a few weeks later, he was regularly timing out his Claude credits and he hadn’t worked in Excel that week. Claude was going into Excel for him." And in the second workshop, people showed each other what they'd built during the week, and those demos turned out to be as valuable as anything I presented.
That week between the workshops mattered, because it gave people time to try things. And time is the usual objection.
When people say they have no time
No one has time to try everything AI can do, because there's far too much of it. But there's enough time to take one piece of work and go deep on it. So when people tell me they have no time to learn AI, I ask for something smaller. Pick one workflow you want to change, and spend time on it over a week or so. I've written separately about which work to start with.
Going deep is what makes that week worth it. It shows you how far you can push the AI, how specific it can get and how much it can improve the result. Ask it for a hundred things at a basic level, and you get a hundred basic results, and none of them will make you want to go deeper.
And the depth carries over. Once you've seen what AI can do with one workflow, you can see how the same approach would work on the next one. The gains build on each other over time, and the workflows that hold up can become agent skills, written instructions the AI follows the same way every time, that the rest of the team can use.
Some people on your team are already well down that road. They're your power users, and they can help bring everyone else along, but what they share matters.
What to ask of your power users
Start with what you're aiming for. The goal is people using AI well, and how much someone uses it doesn't tell you that. Some heavy users use AI poorly, and some people who barely use it use it well.
So when I have power users in a session, I ask them less about what they've built and more about the mistakes they made along the way. People who rarely use AI are often put off by its mistakes, downsides and risks. The power users ran into the same things and pushed through them to get to the value on the other side. Hearing how they got past those roadblocks shows everyone else that using AI well is a skill people develop, and once they see that, they start to believe anyone can do it, themselves included.
Prompt libraries and shared skills
The other way leaders try to spread what their power users know is a shared prompt library, a collection of the instructions people have typed into AI tools. I don't think it solves much. Marty Cagan, the product leader, wrote recently that process is too often used as a substitute for thinking, and a prompt library can become exactly that kind of process. People pick a prompt, copy and paste it, and never think through what the work in front of them needs. And when much of what's in the library isn't good, people can't tell which prompts to trust, and the weak ones produce poor results.
What works better is a library of ideas. Teams share the problems they've solved with AI and how they went about it, and others use that as a starting point for their own conversation with the AI. At Flare, the team set up its own Slack channel for sharing what people were building, without anyone asking them to.
Agent skills can be seen differently. Once a team has built one, tested it thoroughly and knows it does the work well, it becomes an asset the whole company can use. Applying the brand guidelines to every presentation is a good example, since every team that makes presentations can use it and none of them has to build their own. So share agent skills, but only the ones that have been through that kind of testing.
Where to start this week
All of this comes back to people seeing AI work on their own work, and you can start there this week.
- Show your team one example of AI doing a real piece of their work.
- Ask each person to choose one workflow of their own that they want to change, and give it a week.
- When your power users share what they know, ask them about their mistakes first, before showing the final, well-refined output.
Related: Which work should your team hand to AI first? · You've been handed the AI rollout. Where to start with your team · Can your team trust what AI gives them? How to check it · Should your team do an AI workshop or an online course?