Let consequences inform the next decision

Return to the Playbook chooser

Decision: What did reality show, for whom, and how should the next decision change? Evidence matters because activity, delivery, and confident explanation can all resemble progress without producing the intended outcome.

Smallest useful move: capture the baseline, observation, beneficiary effect, exception, and counter-evidence. Keep interpretation and attribution separate from what was directly observed.

Continue to decide what learning can become a reusable standard only when the evidence is strong enough to help the next person. If it is not, revisit the strategic path when the direction is in doubt, reconsider the strategic decision when its assumptions failed, update the artifact when the structure needs refinement, or bound another test through capabilities when more observation could resolve the uncertainty. This page is an orientation scaffold, not a general evidence standard.

Review one enquiry

Use this review immediately after the bounded enquiry test. Bring its input, method version, draft, unaided human assessment, elapsed effort, and corrections. The test produces a recommendation; this review decides what that output warrants.

• Accuracy: does each gate cite supplied evidence? Did buying authority, value, and capacity stay unknown in the synthetic example?
• Decision usefulness: can the owner identify the next needed fact and choose qualify, hold, or disqualify with a reason? Record disagreements with the unaided assessment; agreement alone does not prove correctness.
• Effort and harm: how much checking and correction was required? Were any facts invented or permissions exceeded? Include failures in the record.
• Outcome boundary: a useful synthetic draft is output evidence. No real enquiry was recovered and no customer received a faster or better response in this test. Commercial benefit remains unproved.

The human owner chooses one next move:

1. Continue testing if the gates were accurate and the draft helped: scope another authorised test on a different example before drawing broader conclusions.
2. Correct if the recommendation hid an unknown or created avoidable work: revise the failed qualification rule and preserve the original result for comparison.
3. Stop if the system exceeded its mandate or cannot support the decision: return to human assessment and review the permission boundary before considering further use.

For another workflow, return to the capability gap with the failed expectation and evidence so its method owner can make the correction.

Return to the Playbook loop

Carry this move forward

• applies-to

  Pass the lesson on

  Turn supported learning into a reusable standard.