Good AI training for employees is built around your own tools and your own work, starts each person at the level they're really at, and leaves them with a daily habit and something they keep using afterward. Most AI training doesn't stick because it's the opposite: a one-time lecture or a generic tools demo, delivered to everyone at once.
We run AI training onsite in Cedar Rapids, Iowa City and across Iowa, and virtually anywhere, for staff and for leadership teams. This post covers why so much training fades by the following week, what we do differently, and how to pick a format. The details of what we offer are on our AI training page.
Why doesn't most AI training stick?
Plenty of companies have bought AI licenses, run a lunch-and-learn, and watched usage settle back to the same few people within a month. (If that's where you are, start with our post on AI licenses nobody uses.) The training usually wasn't bad. It was built for the wrong goal.
Four patterns show up again and again.
It's one-to-many. A single lecture to the whole company reaches the people who were already close to the trainer's own level. Beginners get lost, and the people who already use AI every day hear things they knew. One-to-many training usually only lands with one band of the room.
It demos the tool instead of the work. Watching someone summarize a sample document isn't the same as learning to use AI on your month-end close, your intake form, or your project schedule. People don't carry a generic demo back to a specific job.
There's no habit. A two-hour session is a single event. Without something small to do every day afterward, the skill fades along with the memory of the session.
There's no permission or reason to change. If people aren't sure what's allowed, or if working faster only earns them more work, they won't change how they work in the open. A 2025 global study by KPMG and the University of Melbourne, surveying more than 48,000 people in 47 countries, found that only 47% of employees said they had received AI training, and 57% said they hide their use of AI. Training that ignores that second number won't move the first one very far.
What does AI training that sticks look like?
The short answer
AI training for employees sticks when it changes what people do the following Monday, not just what they know. That means six things. Train on the tools your company already pays for, using real tasks from your own workflows. Start each person at their actual level of AI proficiency instead of teaching one pace to the whole room. Give leaders their own session, focused on strategy, policy and how the organization rewards people who improve their work. Build a small daily habit, such as fifteen minutes of learning one new thing with AI's help, so the skill grows after the trainer leaves. Leave people with a takeaway they keep using: a guide, prompts built for their work, or a setup for their own AI assistant. And follow up with coaching or office hours while the new habits are still fragile. A single session can start that. It can't finish it alone.
Levelled by where people really are
We sort people with the AI Proficiency Scale: Beginner, AI Champion and Systems Builder. A level is where someone starts, not a grade. Your best opportunity-spotter may be a Beginner, because they know exactly where the work hurts. Training them like a Systems Builder wastes their time, and training a Systems Builder like a Beginner wastes yours.
AI as a teacher, not only an assistant
The most valuable thing AI can do for an employee is teach them. That's not just a slogan. A 2025 study in PNAS by Bastani and colleagues, with nearly a thousand high school math students, found that students who practiced with an answer-giving AI did worse on a later test without it than students who never had AI, while a version designed to tutor rather than answer avoided that loss. Adults are no different. Training that teaches people to learn with AI builds skill. Training that only teaches them to get answers can quietly erode it.
What formats are there?
Most organizations combine two or three of these. Lengths below are typical, and every format is scoped with you first.
| Format | Who it's for | Typical length | What it covers |
|---|---|---|---|
| Leadership session | Owners, executives, department heads | A focused session, often half a day or less | AI strategy, the difference between AI policy and AI strategy, ownership and incentives for staff |
| Staff workshop | Teams who will use AI day to day | Half a day to a full day | Hands-on practice in your own tools, on your own tasks, levelled by proficiency |
| Proficiency census | Leadership, before training starts | A short survey and conversations over a week or two | Where your people really are, so training starts at the right level |
| Coaching and office hours | Individuals and small teams after training | Short recurring sessions over several weeks | Questions from real work, support while habits form |
| Conference or all-hands session | A whole company, association or event audience | Under an hour to about 75 minutes | A shared vocabulary and one habit everyone can start the next day |
Should leaders and staff be trained separately?
Usually, yes. They need different things.
Staff need practice. Leaders need decisions. Some changes, like security, privacy and compliance, help everyone but no one in particular, so they're set by rule and funded from the top. That's where policy belongs. Most of the value of AI, though, comes from individuals improving their own work, and leaders can't order that into being. They can make it possible by giving people the time to try and a stake in what improves. As we put it to leadership teams: if the only reward for working smarter is more work, nobody changes.
Putting both groups in the same room for a kickoff can help, because everyone hears the same message at the same time. The real work then happens in separate tracks.
Onsite or virtual?
Both work, for different jobs. Onsite is best for hands-on workshops and leadership sessions: people practice together, ask the questions they wouldn't type into a chat window, and the trainer can see who's stuck. We travel for that across Iowa, from Cedar Rapids and Iowa City to the rest of the Corridor and beyond.
Virtual works well for follow-up, office hours, and teams spread across sites or states. A common pattern is onsite to start and virtual to keep it going.
Why does a former teacher run Mosaic's training?
I taught in Iowa public schools for five years before this work, and it shows in how we teach. A good lesson starts from what the learner already knows, gets them doing the work early, and checks understanding before moving on. Terry Doyle's line applies to adults as much as students: whoever does the work does the learning. So in our sessions, people work on their own tasks, in their own tools, most of the time.
You can see how that plays out in a real room in our recap of the AI for project management session at PMI Eastern Iowa, where a mixed room of skeptics, beginners and AI champions each found a place to start.
If you're planning AI training for your staff or leadership team, our AI training page lays out the options, and we're glad to scope a session around your tools and your people. For conference and event sessions, see Speaking & Workshops.