Identify the repeated work and current effort
- Describe the repeated task in one sentence.
- Record its frequency and current preparation and review effort.
- Identify the bottleneck rather than automating the entire process.
- Check whether a form, rule, template, or ordinary integration could solve it.
- Agree which change would make the work meaningfully better.
Check inputs and ownership
- Name the process owner and someone available to test outputs.
- Identify the authoritative source for each required field.
- Collect representative examples, including difficult cases.
- Confirm permissions to read the inputs and use them for this task.
- Separate stable process rules from steps that require model interpretation.
Define permitted actions and failure handling
For an early pilot, a reviewed draft is often enough to test value. Adding write access changes the consequences and should trigger a separate permission and acceptance decision.
- List allowed actions and the tools required for each.
- State which outputs remain drafts until approval.
- Decide how missing data, conflicting inputs, or unavailable systems are handled.
- Check how duplicate requests and repeated processing are recognized.
- Name the person who can stop the workflow and recover unfinished work.
Measure the work from input to approved output
Compare current and proposed work using the same example set. Count preparation time, checking time, corrections, exceptions, and completed tasks. Record source-access failures and denied actions alongside output quality. Do not treat faster generation as a complete measure of value.
Assign maintenance and operating responsibilities
- Who maintains sources, instructions, and integrations?
- Who reviews failures and decides when a change needs retesting?
- Who owns platform updates, credentials, and access reviews?
- What records are retained, for how long, and who can inspect them?
- What result would cause the team to revise or retire the workflow?