APEL (Agentic Product Engineering Lifecycle) is a proposed model for agent-assisted delivery. Work flows through five phases, with clear owners and evidence at each step.
- DiscoverUnderstand the problemProductOUTPUTEvidence and product intent
- RefineMake the target clearProduct · Engineering · QAOUTPUTSpecification and test intent
- BuildImplement and reviewEngineeringOUTPUTReviewed change and evidence
- ValidateTest behavior and acceptanceQA · DevOps · ProductOUTPUTValidation and acceptance evidence
- OperateDeploy, verify, and learnDevOps · QAOUTPUTVerified release and operational findings
The working rules
- Pull when ready. Product sets priority; engineers pull eligible work as capacity opens.
- Specify before building. Carry the intended behavior and test approach into implementation.
- Keep decisions owned. Agents assist; people approve changes, accept outcomes, and control releases.
- Verify beyond the build. Test behavior, confirm product acceptance, and check the live release.
Return to Build with findings. Revisit the specification when needed.
Roll back the release and return the change to Build.
Capture a new issue and feed it into Discover.
Learn and adjust
Compare expected effort with actual effort. Track waiting, rework, and failures to identify what needs attention.
The source model uses on-demand staging and weekly production releases. Choose a cadence that fits operational needs; keep each item’s ownership and evidence visible.
Based on APEL’s September 2026 working draft. This is a proposed model, not a report of measured results.