Review what happens on the second independent attempt
Review the next task completed without live coaching. Compare the accepted output with the intended result, record where help was needed and revise the working instructions around the specific exception encountered.
Ger Perdisatt, Acuity AI Advisory
2 min read
Agree the follow-up before the demonstration ends
Choose the next occurrence of the same type of work and name the person who will try it. Agree what they should save: the inputs they were allowed to use, the output, the checks and a short note about where they got stuck. Use the normal project location and access controls.
Make the intended result specific. For a weekly service update, that might mean an accurate summary of open requests with unresolved items clearly marked. Avoid measuring success through how closely the user reproduced the trainer’s wording. The point is whether the person can complete and check the job.
Record the effort honestly
Keep preparation, generation, checking and correction time separate when timing the task. Record interruptions or unusual source problems alongside the numbers. One slow attempt may reflect missing information, while a fast attempt may have omitted a necessary review.
Use a small number of comparable jobs to understand the pattern. Record whether the result was accepted and what assistance was required. Do not turn one demonstration into a team-wide savings estimate. If the new method adds more checking than it removes elsewhere, investigate the work before recommending wider use.
Editorial guidance informed by Acuity's work with teams. Examples are fictional and client materials stay private. To record who owns an AI use and when it is reviewed, see AI Register.