Illustrative field study

Artificial intelligence

What should AI actually do inside a business?

Give AI a defined role, a safe boundary, and a person who owns the outcome.

Replace the vague request

“We need AI” does not describe a business problem. Start with a workflow people can point to: sorting enquiries, finding a policy, drafting a report, or checking information before someone makes a decision.

Choose the level of responsibility

AI might retrieve information, assist a person, or take a limited action. The more it can change in a system, the more carefully the team must set permissions, checks, and a way to stop or reverse the action.

Test the failure cases

Try incomplete requests, wrong source material, unusual cases, and sensitive data. Decide which outputs need approval and what the system should do when it is unsure.

Measure the work around the model

Track time saved, corrections, escalations, and whether the final work is useful. A model response is not the business outcome. The outcome is a workflow that people can trust and maintain.

Back to all Field Notes