Six months after launch, your team still calls it “the AI project.” That is not a good sign.
When a change really lands, it loses its name. Nobody talks about “the email project” or “the spreadsheet project.” Those tools are simply how work is done. An AI effort that still has a project name, a steering group, and a monthly update is usually not part of the work yet. It is a visitor.
Many companies celebrate the launch and move on. The real test comes later, and it is a quiet one: does the new way of working continue when nobody is watching?
Why projects stay projects
Most AI efforts start well. A small group tries a tool, finds a use, and shows good results. Leaders are pleased. Then attention moves elsewhere.
Without anyone watching, old habits return. Some people keep using the new method, some drift back to the old one, and some never learned it. Results become uneven, and the company concludes that the tool “did not stick.”
The tool usually did not fail. The change was never finished. Nobody decided what the new normal was, wrote it down, or removed the old route.
There is a second reason. The people who ran the pilot were the enthusiasts. They were willing to learn, to experiment, and to put up with rough edges. The rest of the team was never asked to do any of that, so the method has to work for them as well, with less patience.
What normal work looks like
A task has become normal work when four things are true.
• It has an owner. One named person is responsible for how it works and for fixing it when it does not.
• It is written down. The steps, including where AI is used and where a person checks, sit in the same place as every other procedure.
• New people learn it as the way things are done. It is part of onboarding, not an optional extra.
• The old way is retired. Two ways to do the same task means nobody is sure which one is right.
If any of these is missing, the change is still on loan. It will last only as long as the enthusiasm of the people who started it.
Notice that none of the four marks involves technology. All four are management choices. That is good news, because the fix is within reach of any team lead.
A small example
Take a team that uses AI to draft weekly customer updates, with a person checking each one. The plan: 25 updates a week, each taking about 4 minutes instead of 12. That saves 8 minutes each, or about 3.3 hours a week. These numbers are illustrative.
Three months in, the results are mixed. Two writers use the draft every time. One sometimes writes from scratch, because they do not trust it. A new hire was never shown the method and writes everything by hand.
Count it. If only 15 of the 25 updates follow the new method, the team saves 15 × 8 = 120 minutes, or 2 hours a week, not 3.3. The gap is about 1.3 hours a week, or roughly 65 hours a year. The tool did not lose those hours. Habit did.
Closing that gap needs no new technology. It needs an owner, a written procedure, a place in onboarding, and a clear end to the old method.
Five steps to make it normal
1. Name an owner. Pick one person who is responsible for the new way after the project team moves on. Make it part of their role, not a favor.
2. Write the procedure. One page in plain words: what the task is, which steps use AI, what the person checks, and what to do when the result looks wrong.
3. Put it in onboarding. New hires should learn the method in their first week, from the same document everyone uses.
4. Retire the old way. Announce the date after which the old method stops. If a task has two accepted methods, people will choose the familiar one.
5. Review once a month. Check three numbers: how many tasks follow the new method, the time per task, and how often the output needed correction.
None of this needs a big budget. The owner’s time, one page of writing, and a few minutes in each onboarding add up to very little next to the hours at stake.
Why announcing is not enough
Leaders sometimes assume that explaining the change once is enough. People hear the announcement, nod, and go back to their desks, where the old way is still easier.
Habits follow convenience. If the new method is harder to find, harder to start, or not expected by anyone, it will lose. Make the new way the easy way, and make the old way take more effort than it is worth.
This is also why the monthly review matters. It is not policing. It is interest. When a manager asks how the new method is going, people keep using it.
Listen to the people using it
Some resistance is sensible. People worry about errors, about being judged by a tool’s output, or about being replaced. Ignoring those worries creates quiet workarounds.
Ask the team what is not working, and fix the real problems. A draft that is often wrong should be improved or dropped. Trust comes from a tool that keeps its promises, not from a better announcement.
You do not need a survey. Ask three people who use the method every week: what do you skip, what do you fix by hand, and what would you change?
When to call the project finished
Set a finish line at the start. For example: the method is used for at least 90 percent of tasks, time per task has improved by the amount you planned, and the owner has run it for two months without help from the project team.
When those are true, close the project. Drop the name, end the steering meetings, and let the work be work. That is not the end of improvement. It is the moment the improvement becomes part of how the company runs.
Keep a short note of the results at close, such as hours saved per week and the number of tasks covered. It protects the work if budgets are reviewed later.
A project that never ends is not a sign of ambition. It is a sign the change has not landed.
Try it this month
1. List the AI-assisted tasks your team started in the last six months.
2. For each one, check the four marks: owner, written procedure, onboarding, and a retired old way.
3. Pick the task with the most missing marks and fix those first.
Then measure again in a month. If the time per task has improved, you have closed part of the gap. If it has not, find out which of the four marks is still missing.
Do not start new projects until the old ones have landed. Five half-finished changes cost more than two finished ones, and they make people tired of every change that follows.
An AI project succeeds the day people stop calling it one.
Which AI-assisted task on your team has quietly gone back to the old way, and would anyone notice?
If you want to talk through how to make a change stick in your team, I’d be glad to hear your thoughts.
ADDITIONAL READING
• How to Lead Change at Work: Why Announcing a New Process Is Not Enough — https://jordanimutan.com/2026/10/31/how-to-lead-change-at-work-why-announcing-a-new-process-is-not-enough/
• Stop Counting AI Users. Start Counting Hours Saved on One Task. — https://jordanimutan.com/2026/10/29/stop-counting-ai-users-start-counting-hours-saved-on-one-task/
• You Trained Your Team on AI. Their Output Didn’t Change. — https://jordanimutan.com/2026/09/15/you-trained-your-team-on-ai-their-output-didnt-change/
• The Real Reason Most AI Projects Stall Before They Start — https://jordanimutan.com/2026/09/11/the-real-reason-most-ai-projects-stall-before-they-start/
• Why Your Next AI Win Doesn’t Need a Developer — https://jordanimutan.com/2026/10/08/why-your-next-ai-win-doesnt-need-a-developer/
