Only 13% of workers say their employer actually rewards them for redesigning how they work with AI. Meanwhile, 65% say they worry about falling behind if they don't change.
So we've built a workforce that wants to move and an organizational structure that's making it hard to do so. Bit of a design flaw.
Here's the thing about AI investment right now: almost all of it is pointed at the individual. More licenses. More training. More prompting guides. Which makes sense, until you look at what Microsoft recently found.
The 2026 Work Trend Index (20,000 workers, 10 countries, trillions of Microsoft 365 signals) found that organizational factors like culture, manager behavior and how work is designed drive roughly twice the AI impact of individual mindset and skill.
I recently signed my daughter up for piano lessons after she expressed interest in learning. We bought a $50 keyboard off Facebook Marketplace, found a teacher in a nearby neighborhood for weekly lessons, ordered the lesson books, and then stood back to watch the magic happen.
You already know where this is going.
Buying a keyboard and signing them up for lessons doth not a piano player make.
As I was suggesting (pleading?) for the 117th time that she sit down and practice, something occurred to me. Every Thursday, she marches confidently up to the door of her piano teacher as if she has been practicing all week (love her for this). She shows up. She completes the lesson. The "metric" looks great. The habit does not exist.
It's the same with AI at work!
Licenses go out. A training module launches. Completion rates look good. Someone marks adoption as "on track." But three months later, usage has plateaued; slop is showing up everywhere, and another wave of training gets funded.
The cycle starts again.
Training definitely does important work; it introduces a tool and shows people what's possible. But it's usually one moment in time.
Changing how someone works on a Tuesday afternoon when they're under deadline and falling back on familiar habits takes more than a moment. It takes the right nudge, in the right workflow, at the right time. Repeatedly.
And live, instructor-led training can only reach each person once, becomes outdated fast and has no mechanism for reinforcement.
You can't build a habit on a single session.
Here's what I've figured out from our piano situation: dropping her off at lessons and waiting in the car was never going to build the habit at home. I've had to learn a little piano myself. Enough to sit with her at the bench. Enough to recognize when she's stuck on something. Enough to make practice feel like something we do together rather than a chore she does alone.
Once I got involved that way, her practice actually changed.
The same dynamic shows up in Microsoft's data. The 16% of AI users (Frontier Users) who are actually changing how workflows tend to work in organizations where managers visibly use AI themselves. When managers model it, their teams show a 17-point lift in perceived AI value, a 22-point lift in AI-supported critical thinking and a 30-point lift in trust toward agentic AI.
The involvement changes what adoption looks like. You can't sit it out and expect the habit to form.
Microsoft calls them Frontier Firms; they're organizations where individual AI capability and the structure around it are reinforcing each other. Most companies are stuck somewhere in the middle: agents deployed, licenses purchased, training running, results murky.
The ones crossing over aren't necessarily spending more. They've just shifted what they're spending on. Sustained reinforcement instead of one-and-done events. Adoption analytics tied to actual workflow behavior, not just completions. And managers treated as active participants in the motion, not spectators to a program their teams are going through.
I'm not going to pretend there's a magic answer here, but after watching a lot of rollouts up close, a few patterns that show up consistently in the organizations making real progress:
The organizations where adoption is actually sticking are the ones where leaders are visibly using AI themselves — not just sponsoring a program. That 30-point lift in employee trust doesn't come from a top-down mandate. It comes from a manager who can genuinely say "here's how I used it this morning."
The nudge needs to show up at the moment someone is making a decision, not in a scheduled session three weeks earlier. The closer the reinforcement is to the actual work, the more likely the behavior changes.
Completion rates tell you who sat through it. Behavior analytics tell you whether anything changed. Those are very different numbers, and in most organizations they diverge pretty fast. If you're only tracking completions, you're tracking the less interesting one.
Most organizations have employees who are already using AI on their own — sometimes in ways IT doesn't know about yet. That's not a compliance problem. It's a signal. If you want to understand what's actually happening, shadow AI is a good place to start. The companies making the most progress treat it as a roadmap: who's already doing this, what are they using it for, and how do we build around that energy rather than against it.
The best adoption programs don't feel like programs. They show up in tools people already use, at moments when they'd naturally reach for help, and they don't require anyone to leave their workflow to complete something. The more adoption looks like just how work gets done, the more it actually becomes that.
If you want to go deeper on any of these, we've written a lot about reinforcement models, adoption analytics and what behavior change actually looks like:
And if your employees are showing up to their AI "lessons" with the same confidence as my daughter every Thursday, maybe it's time to look at what's happening the other six days of the week.
Connect with the author on LinkedIn: Debra Andersen Wilson | LinkedIn