01 / Section
Start with the recorded attempts
An automation review is only as useful as its account of how work happens. Begin with captured batches that show evidence-linked attempts at a business outcome. Keep the record specific: what was captured, which steps appear connected, and where the evidence stops.
A citation helps a reviewer trace a claim back to its source. It establishes provenance, not certainty. The record may support an interpretation without proving that every step was necessary, typical, or complete.
02 / Section
Compare attempts, not idealized workflows
A single attempt can make an unusual route look like the standard process. Compare distinct attempts at the same outcome and look for recurring steps, differences, and exceptions. Repeated patterns can help identify where a review should focus, but they do not by themselves establish a universal rule.
Keep app and site sequences separate as discovery signals. They can suggest where to look for a workflow connection, but they are not proof that a particular business task happened or that a handoff is ready to automate.
03 / Section
Label evidence, interpretation, and assumptions
Make the review legible by distinguishing three things: recorded evidence, the interpretation drawn from it, and planning assumptions used to consider a change. For example, a captured sequence is evidence; calling it a recurring handoff is an interpretation; assuming an exception can be handled manually is a planning assumption.
Write assumptions where reviewers can challenge them. Note what remains unknown, what would change the proposal, and what should be checked before a decision. This prevents a plausible explanation from quietly becoming a claimed fact.
04 / Section
Leave the decision with people
The purpose of review is to help people decide whether a workflow is suitable for automation, what safeguards it needs, or whether more evidence is required. Reviewers should weigh business context and consequences that captured evidence may not show.
Treat any proposed automation as a decision to assess, not an outcome already achieved. Do not infer employee performance or verified working time from workflow traces, and do not claim savings without separate support. A clear review gives operations leaders a grounded basis for judgment while leaving approval and responsibility with people.
