Look at where the tool sits in the working day, not whether people have a licence. Follow one real task and count what someone has to find, copy or explain again.
Didn’t people just need more training?
Maybe. First find out what happened on the last busy day, when using it was optional and time was short.
1. Follow one task and count the extra steps.
Start. What reminds someone to use the tool?
Input. What must they find, copy or explain again?
Result. Where does the output go, and who still has work to do?
Practical check: Adoption is a working-day question.
A fictional example: change the fit, not the whole toolset.
- Before
- A coordinator opens a separate chat and copies order details to prepare an update.
- Try
- Prepare a draft from approved order information, where the coordinator already reviews messages.
- Keep
- The person checks the details and sends the update.
Practical check: A draft in the right place may be enough.
2. Fix the actual point of friction.
Nobody remembers. Put the step beside the activity that needs it.
The output needs rewriting. Narrow the task or improve the approved examples.
The task isn’t useful. Remove it. Don’t train people to do unnecessary work.
3. Watch one colleague complete one real task.
Ask what they skipped, copied, corrected or couldn’t find. Change one obstacle, then check whether the revised process is genuinely easier to use.
Practical check: A licence count doesn’t tell you whether the work got easier.
Read more: How to Get AI Ideas From the People Closest to the Work · Where to Start with AI: Pick a Problem, Not a Tool
The example is fictional. This is general guidance, not advice for your situation.