The OCEA prompt method
The OCEA prompt method
Process In uses OCEA: Outcome, Context, Constraints, Evidence, Approval. It is a briefing sequence for accountable work, not a magic phrase. Each part removes a different source of failure.
| OCEA element | Question to answer |
|---|---|
| Outcome | What exact artifact and decision should exist? |
| Context | Who is it for, why now, and what happened before? |
| Constraints | What limits, format, tone, scope, and exclusions apply? |
| Evidence | Which sources are authoritative, and how should gaps be marked? |
| Approval | What must stop for human review before action or publication? |
Write OCEA in that order, then inspect it backwards. Approval protects the boundary, Evidence grounds the claims, Constraints shape the work, Context makes it relevant, and Outcome confirms that the job is worth running.
A strong prompt does not remove judgment. It places judgment where the owner can see and exercise it.
Workshop: The OCEA prompt method
The practical objective of this chapter is making outcome, context, constraints, evidence, and approval observable. Before opening a chat or Work, write down how the job is performed today, who owns it, and what counts as an acceptable result. Choose one example you can personally inspect. This baseline prevents the novelty of the tool from being mistaken for real improvement and gives you a fair comparison for elapsed time, accuracy, corrections, and usefulness.
Worked example: a service lead rewrites a vague request for a customer report into a six-section draft with named sources and a human-send boundary. The team first narrows the outcome and assembles this pack: the requester's decision, audience, approved source set, forbidden claims, and named approver. It then runs the agent with an explicit stopping condition, checks every material claim, and records corrections. The example succeeds only when a named owner can explain why the result was accepted, which parts remained human work, and exactly what should change on the next run.
The most common misapplication is adding decorative detail while leaving the outcome or approval boundary ambiguous. The correction is not automatically a stronger model. First reduce scope, rank sources, reveal hidden constraints, and add an approval checkpoint. If the problem remains after those changes, test a higher tier or effort setting on the same sample. This allows you to compare cause and effect instead of merely comparing two different-looking outputs.
- Describe the current human process in five sentences.
- Prepare an evidence pack that includes the requester's decision, audience, approved source set, forbidden claims, and named approver.
- Set one measurable acceptance rule and one prohibited action.
- Ask the reviewer to classify errors by severity, not only style.
- Save the brief, settings, output, and corrections as one run record.
Continue reading the complete book
Unlock all frameworks, permanent library access and future content updates.