How do you know which AI work is worth doing?

Buying a tool can feel easier than defining the work

A new tool gives an immediate sense of movement. The harder question is what should improve, for whom and under which constraints. Without that answer, a subscription or prototype can become another place for the owner to maintain context.

Map an actual journey

Follow one enquiry, job or decision from its starting point to the result. Note the information used, the people involved, the handoffs and the points where work stalls. Ask the owner and the people doing the work to confirm the map. Their differences are useful evidence.

Run a narrow test

Set a baseline before testing. Record completion time, corrections, interruptions or another measure that reflects the outcome. Give the test one owner and a clear stopping point, with a human exception owner before considering automation. A small test can show whether the constraint is information, process, software or capability.

Choose the simplest fit

Compare a process change, existing software, targeted search, AI and specialist support against the same need. Review the result with the people who will use it. Expand only when the evidence supports a wider scope.

Dave can help set the priority, lead a bounded prototype or direct specialist work around an agreed result.

A decision brief keeps the test honest

Write the result in observable terms. For example, an illustrative test might ask whether a service coordinator can find the correct customer context in two minutes with fewer corrections. Name the source information, the person who checks the result and the baseline you will compare.

A useful test can fail cleanly. If the result is not better, you have learned whether the issue was missing information, an unclear process or the wrong intervention.

• Outcome: what changes for the business?
• Workflow: where does the work start and finish?
• Baseline: what time, quality or rework is visible now?
• Owner: who checks the result and decides what happens next?

Review before you scale

Ask the people doing the work what became easier and what became harder. Check edge cases, correction time and the cost of keeping the change current. Only then decide whether to refine the workflow, expand the test or stop.

Common questions