Start here
Collect two or more approved examples of a candidate workflow when possible. Record its trigger, frequency, active effort, error consequences, variability, and available APIs or tools.Copy this prompt
Use this in Screenpipe chat or an assistant with the required connections. Replace the brackets with your details. If the assistant lacks access, provide a reviewed excerpt; a prompt alone does not connect it to your history.Check the result
Test the selected step and measure actual review effort and error rate. Recalculate the estimate with observed results. Stop or revise when correction work exceeds the benefit.See a worked example
See a worked example
A fictional task performed 20 times per week at 6 minutes each has a 120-minute manual baseline. If a tested draft takes 2 minutes of human review per run, the potential gross reduction is 80 minutes per week. A 4-hour setup would take three such weeks to recover before maintenance and failures. These are illustrative inputs, not product performance claims.
Go deeper: setup choices and edge cases
Go deeper: setup choices and edge cases
Good early candidates have consistent inputs, a limited action scope, and an output a person can inspect. A draft checklist or classified queue can be easier to evaluate than a workflow that immediately modifies financial or customer records.Include exception frequency and recovery cost in the comparison. An automation that succeeds on the normal path but duplicates a submission during retries can cost more than it saves.