Founders
How to build an evaluation dataset for AI workflow automation
AI workflow automation: Build an evaluation dataset around the decisions the product must support.
Build an evaluation dataset around the decisions the product must support. Include ordinary cases, difficult cases, and known failure modes with documented provenance. Keep development examples separate from the examples used to judge a change.
Start with the decision
Define the unit being evaluated and the expected result before selecting examples. Record who labeled them, what disagreement means, and how the sample relates to actual use. Revisit coverage when the product or customer workflow changes.
The illustrative scenario
This is an educational workbook. The company or fund and all numerical inputs below are fictional. They are not ADIN results, customer claims, market benchmarks, or a forecast.
The fictional product routes purchase requests through a small operations team. A requester supplies context, a budget owner approves spending, and an administrator maintains the workflow. Exceptions are common enough to affect adoption.
Automating a step can move work to the exception queue rather than remove it. Keep proposed actions, approvals, and completed actions separate in the event history.
An evaluation sample with a denominator
Imagine a permissioned evaluation set with 27 cases and 18 that meet a rubric defined before the test. The resulting percentage describes only this invented sample. Seventy percent is not a recommended target, a safety standard, or evidence about any real system. What matters is the rubric, sample coverage, and consequences of the remaining cases.
Automating a step can move work to the exception queue rather than remove it. Keep proposed actions, approvals, and completed actions separate in the event history. Break the 9 remaining cases into failure categories before interpreting the aggregate. A recoverable wording issue and an unauthorized action may require different product behavior even when both count as a failed case. Keep reviewer disagreement and missing evidence visible.
The assumed time comparison is 18 minutes before and 11 minutes after, including the stated human step. Validate the same start and end points in a real study, then include retries, repair, and handoffs. Compare the full request-to-decision process with the current system. A faster initial classification can hide a slower approval or correction stage. A short average time is not enough if the result cannot be used or trusted.
| Item | Illustration |
|---|---|
| Invented evaluation cases | 27 |
| Cases meeting the fictional rubric | 18 |
| Cases requiring investigation | 9 |
| Observed proportion in this exercise | 66.7% |
| Assumed human minutes before / after | 18 / 11 |
Use this worksheet
Answer the questions with a source, date, owner, and next action. Mark an unknown as unknown. Use public or permissioned information and involve a qualified specialist when a question requires expertise beyond the team’s competence.
- What is the evaluation unit?
- Where did the examples come from?
- Who labeled the expected result?
- Which failures are underrepresented?
- What data must remain held out?
Questions specific to this scenario
- Who approves a purchase?
- Which exceptions stop the workflow?
- How are manual overrides recorded?
Inspect request histories, approval timestamps, exception reasons, manual overrides, and changes in process ownership. Include requests that stalled or were withdrawn.
A concrete next test
Follow one request that passes normally and one that needs an exception. Ask the requester, approver, and administrator what each can see and change. Record whether automation removes work or increases the burden on the person managing stalled requests.
Interpret the result
A larger dataset is not automatically more representative. An evaluation should reveal consequential behavior in the intended workflow. If repeated tuning teaches the team the answers, add independent examples before claiming a general improvement.
Compare the full request-to-decision process with the current system. A faster initial classification can hide a slower approval or correction stage.
Sources and method
These primary resources provide background for the concepts. The fictional scenario, calculations, and worksheet are ADIN educational examples; the sources do not report or validate those invented results. Reviewed October 11, 2026.