Training your team to work with an AI feature

A serial entrepreneur with a finance background, who has launched and exited several startups, corrected me gently when I called his track record a string of successes. Some of those, he said, were learning experiences. He was honest that the resume shows the highlights and hides the living-through-it part that produced them. Then he described how he actually gets better. At a big startup event, he said, you do not go to win the award. You go to be around incredibly capable people doing hard things, and you hope that through osmosis you become a little more like them, and maybe help them become better too.

That is the truest description of AI training I have heard, and it is nothing like how most companies do it. Most companies train people on a new AI feature with a launch email and maybe a recorded walkthrough. Then they are surprised when adoption is shallow. Nobody got fluent at anything from an email. People get fluent through reps, through proximity to someone further along, and through a couple of learning experiences that stung enough to teach them.

A briefing is not training

Around 38% of teams name a lack of role-specific training as a barrier to AI adoption. But the word “training” hides the real gap, because most of what fails was technically delivered. There was a session. There were docs. What there was not, was reps and proximity, the two things the entrepreneur named. You cannot brief someone into fluency any more than you can read your way to knowing how to ride a bike.

His osmosis point is the practical core. Fluency spreads from people, not from documents. Put a new user next to someone who already works well with the AI and they absorb the habits in a week, the small moves that no walkthrough captures. Isolate them with a PDF and they stay beginners. The learning experiences matter too. The person who has already prompted something badly and watched it go sideways knows things the briefed person does not. That is why he goes to events to stand near capable people rather than to collect a prize. Proximity teaches. A certificate does not, and neither does a slide that says the rollout is complete.

What training that works looks like

Reps before theory. Have people do a real task with the AI on day one, not watch a demo of it. The doing is the training. Everything else is context.
Pair the new with the fluent. The entrepreneur’s osmosis is real. Sit a beginner beside someone further along and fluency transfers through proximity, not paperwork.
Let the small failures happen safely. A learning experience that stings a little teaches more than a perfect walkthrough. Build a space where a bad output is a lesson, not an incident.
Train the role, not the tool. A generic AI overview helps no one. Train the specific person on the specific task they own, or you have delivered a briefing and called it training.

How we approach it at Density Labs

Our AI Readiness Assessment ($2,500) plans the training as part of the rollout, because we have seen too many good features stall on a launch email. We identify who needs to become fluent, who they can learn next to, and which real tasks give them their first reps. Training that is a document gets skimmed. Training that is reps plus proximity actually changes how people work.

You cannot email someone into fluency. Put them next to someone further along, give them a real task, and let a few of the reps go wrong. That is how anyone learns anything.