Interest is not the same as changed work
A good training day can make people curious. It can show what a tool can do and give the team common words.
The change often fades when everyone returns to a full diary. The examples were safe. The data was clean. No real customer, deadline or hard call was involved.
Practice changes more because it meets the work where it is messy.
Pick a job with a real owner
Choose one job that matters to one named person. It should happen often enough to test again soon. It should also have a result that can be checked.
Good examples include:
- preparing a weekly sales view;
- turning customer calls into product questions;
- checking a proposal against the buyer's needs;
- bringing several reports into one short brief.
Avoid a broad goal such as “help the team use AI”. It has no clear finish.
Keep the before and after
Save one recent example of the work before AI was added. Note the time, the source files, the mistakes and the person who checked it.
Build a small first version. Use it on live work. Then compare:
- Did it return useful time?
- Did the work become easier to check?
- Did the person keep the final call?
- Could someone else understand how it works?
This is proof the team can discuss without guessing.
Teach through the change
Ask the owner to explain three things to a colleague:
- what the AI carries;
- what the person still decides;
- what makes the system stop and ask for help.
If they cannot explain those points, the work is not ready to spread.
What should remain after the month
The useful result is not a folder of prompts. It is a working job, a clear owner, a few good examples and a record of what was learned.
Train on a real result. Let the lesson come from the work.