Lowy Financial journal

A useful task is a good starting point.

The clearest description of an AI product often begins with the work it is meant to do.

Ask a team how it handles the task today. Where does it spend time? Which steps cause delays? A concrete answer makes it easier to see how software could fit in.

Take an operations team that sorts incoming documents and routes them for review. A trial should include incomplete documents, unfamiliar formats, and items that belong with a different reviewer. Those cases show when a team member needs to step in.

Set boundaries for the trial

Choose what the tool can read and what decisions a person needs to make. Test with examples that resemble actual work and track how much time staff spend correcting or checking output.

Explore our AI investment focus

Measure how the work changes

Before a pilot starts, record how long the current process takes and where mistakes tend to happen. Compare that baseline with the trial. Count setup and review time alongside the time saved on the task itself.

Back to the journal