A playbook for the rollout — the unsexy change-management work where most AI projects actually die. Training without bureaucracy, champions and skeptics, and the metrics that tell you whether it's actually working.
Most AI projects don't die from a bad model. They die because the team didn't adopt them — quiet resistance, parallel systems, exceptions that became the rule. This playbook is the change-management volume of the Library. Less code; more people; equal stakes.
Adoption is a leadership job, not a tooling job. You can't outsource the conversation about how the team's work is changing.
If you've watched two or three AI rollouts in a row, you've seen these patterns. None are exotic. All are preventable if named in week one rather than week ten.
No senior person owns whether this happens. People sense the lack of weight; effort slips. Pilot drifts; never lands.
Fix: a named senior sponsor with a quarterly outcome. "By Q2, support response time is down 40%."
Unspoken in meetings, dominant in private. Subtle obstruction. People discover edge cases that suspiciously kill the project.
Fix: address it on day one. In writing. With specifics. "This automates X. Your role is Y. Here is what's changing for whom."
Login. Switch app. Wait. Read. Approve. Send. Old way: one keystroke. New way wins on capability; loses on flow.
Fix: measure end-to-end clicks and seconds. If the AI workflow is longer than the manual one, you have a UX problem, not a model problem.
Some people will. Most won't. They'll use 10% of the tool and report it as "fine." Quietly, the leverage doesn't show up.
Fix: 30 minutes of real training per person. Real prompts. Real wins demonstrated. Not a recorded video.
Three months in, you ask. The answer is "I think so." That's not data. The project becomes inertia, then dust.
Fix: a metric per workflow, reviewed monthly, posted publicly. Trend matters more than the number.
"We rolled out the AI." People aren't a deployment. They are colleagues changing how they work. The patterns above are versions of that one mistake.
Fix: act like you're changing a team's daily life, because you are.
A rollout that ignores how people feel is a rollout that ignores reality. Adoption is half technical, half emotional. The half you neglect is the half that kills you.
Change-management at corporate scale is a profession. At small-team scale, it's an attitude — a handful of disciplines applied honestly. Five moves; none complicated; each makes the next easier.
01 · Name the change. A one-paragraph internal memo. What's changing. Why. By when. What's not changing. Read aloud once; available in writing forever.
02 · Name the champions. One or two people on the team who are visibly excited. Not anointed by you — already visible. Give them air cover and air time.
03 · Listen to the skeptics. Not "convince." Listen. Most skeptics are pointing at a real risk; the project gets better when their concerns are heard early.
04 · Show the work. Demo wins weekly. Concrete. Specific. "This used to take Sara two hours; now it takes 12 minutes." Not slide-shaped; case-shaped.
05 · Adjust the memo. Six weeks in, the original memo is wrong somewhere. Rewrite it. Acknowledge what changed.
Roughly: 20% champions, 60% middle, 20% skeptics. Spend most of your energy on the middle — they're persuadable, they're the volume. The champions are already there. The skeptics will move when the middle does.
"This is too soft." "Just ship and they'll see." A founder is usually right that the team will eventually see. They're usually wrong about the cost of the eventually.
Punishing people for the old way while the new way is still half-built. The bridge has to exist before you close the old road. Don't reverse the order.
Adoption work feels slower than building. It is not. It is what makes the building count.
Write the change memo. One paragraph. Share with the leadership before sharing with the team. Refine. Then send. Don't outsource the writing.
Most corporate AI training is a recorded video your team will skip and a quiz they'll game. Small-team training works because it's specific, hands-on, and tied to the actual work. Spend the budget here, not on certifications.
5 min: what AI is (Vol. 01 mental model), what it's bad at.
10 min: show me one task you do that you hate. Try it together.
10 min: they try one. You correct in real time.
5 min: save the working prompt to the snippet library. Done.
A standing weekly 30-minute slot where anyone on the team can bring a prompt, a workflow, or a confusion. Champions show up; skeptics drop in once; middle hangs around. The single most effective standing meeting most small teams add when AI rolls in.
If your team's AI usage spike after training drops to zero a week later, the training was theatre. Real adoption shows up as a flat-ish line of regular use, not a spike.
Training is the conversation in front of the work. Skip the workshop; show up to the work.
Book three one-on-ones with members of your team. 30 minutes each. Bring no slides; bring your laptop. Ask them what they hate doing on Tuesdays. Solve one together.
Vanity metrics make rollouts look successful while they fail. Honest metrics make rollouts look messier than they feel. Choose honest. The team can handle reality; they can't fix a problem they can't see.
30 minutes. The leadership team. Three slides max:
A rollout where one workflow saves 8 hours a week for 5 people, and four workflows are abandoned, is a successful rollout. A rollout where five workflows are "in use" with no measurable time savings is not. Concentration of impact beats breadth of usage.
Don't ask "do people like the AI?" Ask "is anyone's Tuesday quietly different than it used to be?" Different Tuesdays are the only real metric.
For each workflow you're rolling out, name the metric, the baseline, the target. Write all of it in one place. Add the monthly review to the calendar. The discipline is in the recurrence, not the dashboard.
Two months from "we should do something about AI" to "the team has measurably adopted one or two workflows." Realistic for a team of 5–25. Scale the days, not the structure, for larger teams.
AI Adoption for Small Teams · The Operator's Library · No. 12. The final volume of the Library's planned 12.
Adoption is not a project. It is a programme — small, recurring, social, honest. Run it like the leadership work it is.
People before tools. Champions before skeptics. Memo before metrics. Tuesdays before quarters.
— END · OPERATOR'S LIBRARY NO. 12 · LIBRARY COMPLETE