Which Problem Deserves an AI Pilot?
A team asks for AI pilot ideas. Three proposals come in. The team picks the one it can demo by Friday.
The demo works. On Monday, everyone goes back to the old workflow.
The gap
The team knows which idea is easiest to demonstrate. It has no shared criterion for deciding which problem is worth solving.
The provocation
PO — An AI pilot succeeds when something disappears.
From provocation to possibility
Imagine the pilot succeeds and one step disappears from the weekly workflow. Which step would the team actually be relieved to lose?
A customer support team might choose to stop sorting every incoming ticket by hand. That gives the pilot a specific purpose. Can AI handle routine sorting well enough for staff to spend their time reviewing uncertain cases?
Try it on 20 tickets
- Choose 20 recent tickets that include both clear and ambiguous requests.
- Agree on the correct routing and identify any routing errors the team considers unacceptable.
- Compare manual sorting with AI-assisted sorting on the same tickets. Include the time staff spend checking and correcting the AI’s suggestions.
- Use the results to decide whether the workflow is ready for a larger trial.
Before choosing an AI pilot, ask which step you’d be willing to retire if it worked.


No comments:
Post a Comment