AI rollouts fail because they start with the tools and ignore the people who have to use them. MIT's NANDA initiative studied hundreds of enterprise deployments for its 2025 report, The GenAI Divide, and found that 95 percent of generative AI pilots deliver no measurable return. The report puts the blame on a learning gap inside organisations rather than on the quality of the tools.
That finding should change how a small business plans a rollout. If adoption is a people problem, the fix starts with people, and it looks less like a software launch and more like a change in ways of working.
What the research actually says
Three findings from The GenAI Divide matter most for smaller teams. Only around five percent of AI pilots reach production with measurable impact. The barrier the researchers name is organisational, a learning gap between tools and the businesses using them, rather than any shortfall in the models. And more than 90 percent of workers already use personal AI tools, even in businesses where the official rollout has stalled.
That last one deserves a moment. Your people are probably already using AI, quietly, on their own terms. A rollout that ignores this fights the current instead of working with it.
The people problems a rollout meets
- Confident people race ahead, build private workflows, and stop sharing what they learn.
- Nervous people go quiet, avoid the tools, and hope the whole thing passes.
- Nobody agrees the rules, so everyone quietly decides for themselves what AI is allowed to touch.
- Training happens once, on a busy day, and has faded by the following Friday.
None of this shows up on a software dashboard, which is why so many leaders only discover it at renewal time, when the licence count and the usage numbers finally meet.
Start with how your people actually work
Different brains meet new tools differently, and a rollout that works for every brain starts by finding out where each person honestly stands. That is why the Working With Me method begins with a Working Profile, each person's own working manual, which for AI adoption also captures their AI confidence and the support they need.
The point is honesty rather than assessment. Nobody is scored and nobody is watched. The profile tells you who wants to run experiments, who needs reassurance and time, and who sits somewhere in between, so the plan can meet people where they are.
Agree the rules of the game
Most AI anxiety is really uncertainty about permission. So the team agrees the rules of AI-enabled work together: what AI is for, what stays human, and what gets checked before it goes out. Written down where everyone can see them, the rules replace a hundred private guesses, and the nervous people finally know where the ground is.
Turn training into a weekly rhythm
A training day teaches what a tool can do. A rhythm teaches a team what the tool is for. Each week the team tries one small experiment, shares what worked at the check-in, and keeps what earns its place. Experiments that add nothing get dropped, and dropping them counts as progress, because the goal is better work against the priorities that matter rather than more AI use for its own sake.
Show the evidence without watching anyone
Leaders still need to know it is working. Evidence dashboards show the patterns across the team, connected to the priorities, while private answers stay private and nobody is monitored. You can see adoption becoming normal working life, and your people keep the privacy that makes their answers honest.
Where to start
The fastest way to feel this working is the Working With Me workshop, a single day from £2,000 where your team builds their Working Profiles, agrees the rules of AI-enabled work, and leaves with a four-week trial of the platform to run the first experiments. The AI adoption module shows how the whole method fits together.
Common questions
Why do most AI pilots fail
MIT's 2025 report The GenAI Divide found 95 percent of enterprise generative AI pilots deliver no measurable return, and names a learning gap inside organisations as the barrier rather than the technology. In plain terms, tools get bought and people are left to catch up.
How do I get my team to actually use AI
Find out where each person honestly stands first, then agree the rules of AI-enabled work together, then run one small experiment a week and keep what earns its place. Habits form through rhythm, and a team that agreed its own rules has far less to fear.
How long does AI adoption take
Expect a quarter of weekly rhythm before new habits feel normal, and treat the four-week platform trial that comes with the workshop as the first stretch of that quarter. A single training day on its own rarely changes anything by the following month.