Build
From specification to production-ready increment
In the Build phase, multi-agent systems implement the target solution based on the quality-assured specifications from Analyse and Design. Consultants steer the agents, assure quality through reviews and policies, and keep the delivery track's DORA metrics at elite level.
Position of the phase in the TRANSFORM delivery model
Process model
The phase delivers production-ready, tested increments with full traceability from requirement to code — shipped in short cycles over a trunk-based, fully automated integration track.
Delivery setup & definition of done
Set up the agentic delivery track (repositories, CI, policies, environments); agree the definition of done, coding and review standards with the client team.
Specification-driven agent implementation
Decompose epics into agent-ready work packages with precise acceptance criteria; multi-agent code generation against the Design phase's contracts and ADRs.
Continuous integration & review
Trunk-based development with small batches; every change passes automated checks and a human review of functionally critical paths.
Quality & security gates
Static analysis, dependency and secret scanning, architecture conformance checking against the target architecture; policy gates block non-conforming changes.
Measurement & learning loop
Continuous collection of DORA metrics and agent effectiveness; retrospectives to improve prompts, contexts and work package cuts.
Methodology mix
Agentic implementation under human quality authority
| Method | Purpose in the TRANSFORM context | Reference |
|---|---|---|
| Trunk-based development | Small, short-lived branches and daily integration; precondition for agentic delivery speed without merge conflicts. | trunkbaseddevelopment.com |
| Continuous delivery | Deployment pipeline with automated quality stages; every change is potentially releasable. | Humble & Farley (2010) |
| Test-driven development | Test-driven specification of functionally critical logic; acceptance tests as executable requirements for the agents. | Beck (2002) |
| DORA metrics | Deployment frequency, lead time, change failure rate and MTTR as empirically validated steering measures of the delivery track. | Forsgren, Humble & Kim (2018) |
| Clean code & SWEBOK | Binding coding and maintainability standards; SWEBOK as the reference frame of software engineering disciplines. | IEEE CS (SWEBOK v4) |
| ReqPOOL Agentic Engineering Playbook | In-house standard for work package cutting, context engineering, agent reviews and human-agent division of labour. | ReqPOOL (internal); cf. Anthropic (2024) |
Platform support
ReqPOOL Suite: reqCoder (AI software development) · Estimation Manager (Estimate · Scope · Govern)
reqCoder — agentic implementation (coming soon)
EU-based AI develops the software directly from the AI coding intent — no client data used for training, complete audit trail; human engineers review, harden and integrate the generated code.
Specification as Code as working basis
The AI-generated implementation order from the Estimation Manager (screens, use cases with acceptance criteria, business objects, interfaces, permissions matrix) serves as the direct working basis of agentic development.
Estimation Manager — scope steering during build
Scope changes during implementation are captured as versioned change requests; scope drift is continuously measured against the baseline.
Traceability via IDs
Requirement IDs from specification and coding intent link work packages, code and test cases — a complete evidence trail for audits and quality gates.
Platform usage best practices
- Cut work packages along the use cases of the AI coding intent — clear context, executable acceptance criteria, limited blast radius.
- Formulate acceptance criteria as tests before generation; the AI implements against executable requirements.
- Focus human reviews on functionally critical paths and architecture boundaries; never integrate generated code unchecked.
- Capture every scope change immediately as a change request in the Estimation Manager — scope drift is measured against the baseline (gate: code complete).
- Review DORA metrics weekly in the team; treat regressions as impediments, not as reporting figures.
- Version prompts and context packages and improve them in retrospectives — agent effectiveness is an engineering artifact.
Artifacts & outcomes
- Production-ready, tested increments per release
- Full traceability requirement → code → test
- Automated integration and quality track
- Migrated and modernized legacy components
- DORA metric baseline and delivery report
Quality gate — Build
Transition to the next phase happens through a formal quality gate (go/no-go). Gate criteria are documented in the gate review and signed off by the engagement lead.
- All increments meet the definition of done
- Policy gates passed without open exceptions
- Traceability chain proven without gaps
- DORA metrics within the agreed target corridor
Scientific deep dives
Empirical foundation of DORA metrics and high-performance delivery.
Ongoing research on software delivery performance incl. AI impact.
Standard reference on deployment pipelines and release automation.
Foundational work on test-driven development.
Body of knowledge of software engineering.
Design patterns for reliable agentic systems.
Best-practice guide: Build phase
Compact checklist covering process model, methodology mix and platform best practices for use in your engagement.
Download the best-practice guide (PDF)