Joe Cook

Systems libraryDesign & Production

Designing AI Training That Sticks

Why most AI training fades within a month, and the approach I use so a team is still using what it learned 90 days later.

Joe Cook,

Most AI training follows the same pattern. Someone gives an energetic presentation about what AI can do. People leave impressed. A month later almost nobody uses it any differently than before.

The problem usually isn’t the content. It’s that the training was written for nobody in particular and never touched anyone’s actual work. Here’s how I build it instead.

Start with discovery

Before I design anything, I spend 60 to 90 minutes with the people who can make decisions. I want to understand six things:

  • The organization. Size, structure, business lines, and which roles face customers.
  • The people in the room. How comfortable they are with technology, what tools they already use, and how much they’ve already tried AI.
  • What leadership wants. How they’d like staff to use AI, and what they’re nervous about.
  • Priority use cases. Four to six everyday tasks where AI should help first. Not twenty. Four to six.
  • Guardrails. Data sensitivity, IT policy, and which tools the company already licenses.
  • What good looks like. If we’re successful, what will people be doing differently 90 days from now?

That last question shapes everything else. If leadership can’t answer it, the training will drift.

After the meeting, I send two short intake forms. One goes to the organization and asks about participants by department, existing policies, and scheduling. The other goes to each department and asks for the tasks and questions people most want help with. Finance and marketing will give very different answers, and their training should reflect that.

Choose a small set of tools

Part of discovery is settling on the core tools everyone will learn. That might be ChatGPT, Claude, Microsoft Copilot, or whatever the company already pays for. Some departments will need something more specialized, and that’s fine.

What doesn’t work is trying to cover every tool at once. People end up with a shallow sense of five products and real skill in none. Pick a core set, train on it well, and add specialized tools where a department needs them.

Three rules I design around

Every session I build follows three rules.

Context before tools. People hear why AI matters to their own work before they open any software. A salesperson cares about prep time before a customer meeting. An accountant cares about month-end reconciliation notes. Start there.

Practice over lecture. Every session includes hands-on time with scenarios from the job. People bring a real task, work through it with the tool, and leave with something finished. Watching someone else use AI doesn’t build the habit.

Leave with a commitment. Each person picks one skill to use on real work within five business days. Just one. Confidence comes from doing it once successfully, not from hearing about it.

A shared foundation, then department workshops

The structure I use has a foundation everyone takes together, followed by workshops shaped around each department.

Training structure. Discovery leads to a shared foundation for everyone, then a checkpoint, then workshops by department, with on-demand support afterward. Discovery Foundationeveryone Checkpointdecide next Workshopsby department On-demand Narrow from the whole organization to each team's real work.

The foundation is short. A live session covers what AI is, why the organization chose its tools, and gets people using them right away. A self-paced prompting module follows. A second live session a couple of weeks later answers questions, shares what people discovered, confirms everyone has the basics, and introduces AI agents, assistants set up with standing instructions for one recurring job.

Then comes a checkpoint. Leadership looks at how the foundation went and decides whether to continue as planned, adjust, or stop. Nobody should be locked into more training than they need.

The department workshops are where the real change happens. Each one runs about an hour and is built from that department’s intake form. Proposals works on proposals. HR works on job descriptions and onboarding. Each workshop includes a custom agenda, prompts and templates written for that team, two or three applications chosen with them, a hands-on exercise, and a checklist they keep.

Afterward, departments can request on-demand sessions when they hit a new use case or need more time.

What people take back to their desks

Training fades when it lives only in someone’s memory. So every session leaves behind something to use:

  • A prompt cheat sheet. One page of prompts written for that organization’s real tasks.
  • Job aids. A short checklist for each department’s workflow.
  • A participant workbook with exercises, templates, and a personal action plan.
  • Recordings, so new hires can catch up.
  • A self-paced module the company can load into its own learning system.

Name an internal champion

The single biggest predictor I’ve seen of whether training sticks is whether someone inside the organization owns it afterward. That person answers questions, collects good examples, notices where people get stuck, and keeps the cheat sheets current. It doesn’t need to be a technical person. It needs to be someone people trust and will actually ask.

I ask for that name during discovery. If nobody can be named, that tells me something important about how the next 90 days will go.

The full engagement

If you’d like to see how this comes together as an engagement, including a typical timeline, the AI training engagement page lays it out. Every piece of it can be resized for your organization.

Talk with me about building something like this