THE ANSWER
Define one task, agree its acceptance checks and information-handling boundaries, then compare complete work across a representative sample. Record errors and total effort. Finish with a documented decision to continue, change or stop before extending use.
An illustrative four-stage plan
The stages below are a planning template. They are not a delivered client project or a promised timetable. A stage should end when its evidence is ready, rather than because a calendar week has passed.
| Stage | Work | Evidence needed to move on |
|---|---|---|
| 1. Define | Choose one internal output, owner, boundaries and stop conditions. | Written brief, approved information arrangements and review method. |
| 2. Establish a baseline | Observe the existing process using comparable tasks. Separate active time from waiting. | Preparation, drafting, correction and review records. |
| 3. Trial and inspect | Save outputs, include difficult inputs, log errors and maintain review. | A record for every sample, including rejected outputs and corrections. |
| 4. Decide | Compare accepted work and total effort; review unresolved risks. | Named decision, reasons, limits and a next review date. |
Measure the whole task
For each item, record preparation, tool operation or drafting, correction and review time separately. Add those components for total active effort. Keep waiting time in a separate field. Use the same definition of an accepted output for the baseline and pilot.
Report the number and nature of items, what was excluded and whether the same people reviewed both sets. A small or unusually easy sample should not support a firm-wide savings claim. This publication has not measured a human baseline or time saving.
Decide the stop conditions in advance
A proposed starting point is to pause on an unauthorised information flow, unsupported recommendation, automatic external sending outside scope, or repeated critical factual error. Adapt the criteria to the actual task and firm responsibilities.
A pause should preserve the evidence needed to understand the issue. Agree who investigates, who can change the process and what would justify a restart.
Write the decision so someone else can assess it
Record the task and configuration tested, sample and exclusions, time components, error categories, reviewer findings and unresolved matters. Explain why the evidence supports continuation, a changed trial or stopping.
Wider rollout needs its own scope, training, ownership and review arrangements. An internal action-list exercise does not establish that the same tool is ready to draft recommendations or communicate with clients.
APPLY THIS TO YOUR FIRM
Want help turning this into a practical AI workflow?
Work with Steven Calvert on an assessment, a defined pilot or staff training for your IFA business.
Explore AI implementation ↗Prepared with AI assistance. Our examples identify their sources, method and limitations. This article provides business guidance, not personal financial advice, legal advice or a product endorsement. Editorial approach and commercial interests.