What to use AI for when your team draws a blank

You bought the AI tools, ran the kickoff, and a few weeks later most of the team still isn't using them. The easy read is that people lack skill or willingness, but that's usually wrong. The harder truth is that they don't know what to use AI for on their own work, and figuring it out feels like more effort than just doing the task the old way. The data backs this up: a WalkMe survey found nearly 60% of employees say it often takes longer to work out how to use AI than to finish the job manually, and Forrester's research found only about 26% of employees actually understood prompt engineering in 2025, up a mere four points in a year. Idle licenses aren't a sign people are lazy. They're a sign nobody helped them find the fit.

That reframes the problem from a skills gap to a discovery gap, and a discovery gap calls for a different response. You don't close a discovery gap with another tool or a sterner email. You close it by helping each person see where AI belongs in the work already on their plate.

The instinct is to hand the team a list: fifty great prompts, go. It rarely takes, because a prompt written for a generic marketer or analyst misses the specific document, system, constraint, and deadline of this person's actual task. So people try it once, it doesn't quite fit, and they conclude AI isn't for their job. It's why so many users get stuck doing the same shallow things, summarizing an email, a quick search, and nothing higher value. By Section's own research, the overwhelming majority of employees use AI only for basic tasks that generate little return. A list shows people what AI can do in the abstract. It doesn't bridge to what it can do for them, today, on the thing they're actually stuck on.

The more sophisticated version of the prompt list is the use-case catalog: a searchable library of scenarios sorted by role and outcome. It's better than a PDF, and it has the same ceiling. A catalog surfaces ideas; it doesn't tell you whether a given idea fits the specific task you own, and that fit is exactly the gap that makes AI feel like more work. Here's the strongest objection to everything I'm about to argue: maybe you don't need coaching at all, you just need a great catalog, or you let the tool suggest uses. The honest answer is that catalogs get browsed once and abandoned, because there's no accountability loop, which is precisely why seats go idle. The tell is in the market itself: even vendors whose whole product used to be a role-based course library have concluded the catalog wasn't enough and moved to coaching people on their own work. Menus don't create fit. A coach working through your actual task does.

When a team draws a blank, the answer to what to use AI for is not a blank-slate brainstorm of everything AI could theoretically do. It's to start where people already spend their time, with the routine, repetitive, low-risk parts of their own work. OpenAI's own guidance points the same way: the highest-value starting points are the cognitive bottlenecks and recurring chores in workflows you already run, well ahead of novel moonshots. Pick a high-volume, low-risk task, drafting a recurring report, sorting a spreadsheet, cleaning up meeting notes, and get one real win there. That win does two things: it proves the fit on this person's actual work, and it builds the confidence to look for the next one. Anchor the search in each role's real tasks and it stops being abstract.

Finding the fit isn't a one-time workshop; it's a habit, and habits need reinforcement. The teams that get past the blank stare have two things: someone coaching each person through discovery on their real tasks, and a manager who keeps it alive. Managers hold the lever here, and they use it as coaches: asking what people tried this week, celebrating a real win, and helping whoever is stuck, with no usage dashboard involved. That reinforcement separates a team that quietly gives up from one that compounds, and it's how AI adoption actually takes hold across a team. A named champion with manager backing accelerates it further, because discovery spreads fastest from the people closest to the work; a central list can't match that.