What is RAPID?

RAPID — Rigorous AI-Powered Iterative Delivery — is Davies' model for building software at the new speed: AI agents do the building; engineers direct, review and govern.

AI is rewriting how software gets built. Work that took teams weeks or months of expensive effort — research, design, build, test — can now be done by an agent in hours or days at a fraction of the cost. The bottleneck is no longer typing code — it's direction.

RAPID is our answer: vibe-coding-style acceleration with enterprise engineering rigour built in — architecture, security, compliance and test as part of the flow, not bolted on at the end.

The shift: from typing to directing

You decide the direction and authorise the irreversible. The agent investigates, builds, verifies, and captures what it learns. The loop compounds.

Definition → Design → Tech selection → Build → Refinement → Flow-state ⇄ Production

• Before: a rigorous lifecycle was affordable only for big bets — the careful parts (capture, polish, hardening) got cut first.
• Now: every phase gets cheap, so you run the full lifecycle at any scale — even a single bug earns the whole loop — and still capture what you learn.

Where to go next

The story (from the deck):

• The case for change — where we are today and the cost of standing still.
• The AI delivery spectrum — vibe coding vs AI-enhanced development vs RAPID.
• The RAPID model — the pipeline, the loop, and the outputs of every phase.
• Stop · Go · Build — the three-month plan to get there.
• Operating model — squads, specialists and stakeholders.

The process (how we actually work):

• The RAPID process — the master definition of the seven phases, handoffs and scaled variants.
• Gates — the checklists that keep speed safe.

Standards — the rules, written for humans and loaded by agents. Grouped by theme:

• Governance — authorise the irreversible, decision records (ADRs), repo context & memory, lifecycle RACI, toolkit governance.
• Decision-support — prioritisation, reference architectures, model selection, token budgeting.
• Build — coding standards, code review, CI/CD, branching & commits, docs with build, Definition of Done, supply chain, observability, AI provenance.
• Definition & Design — acceptance criteria, domain glossary, NFR catalogue, accessibility, content & microcopy, data-model governance, insurance UX heuristics.
• Quality — test pyramid & coverage, AI-generated-test review, test data, test environments, flaky tests, defect taxonomy, regression.
• Security — agent guardrails, secure by default, security gates, secrets management, LLM app security, AI decision auditability, pen-test cadence.
• Delivery & measurement — metrics framework, remediation backlog, async comms, release comms, rapid-fix SLAs.
• Commercial — client transparency, product telemetry, demo environment.

Playbooks — how the work is done:

• Engine — director–builder loop, tech selection, refinement, flow-state.
• Decision-support — tooling evaluation, golden-task harness.
• STOP methods — tech-debt audit, security-gap assessment, DORA baseline, FinOps review, ways-of-working audit.
• Definition & Design — discovery & research, PO directing agents, design phase.
• Quality & security — test strategy, manual testing, automation testing, performance testing, threat modelling, AI-defect incident.
• People & ways of working — squad playbook, SME engagement, shared specialists, UK–India collaboration, delivery cadence, community of practice, squad onboarding, support & ops staffing.
• Commercial & GTM — feedback pipeline, GTM collateral, client cadence, support, beta programme, KCS knowledge capture.

The operational counterparts — gate cards, templates, the project scaffold, prompts, skills, radar and registers — live in the repo's toolkit/ tree (see toolkit/README.md).

note

RAPID was formerly named RAID (Rapid AI Development); it was renamed to avoid confusion with RAID logs. Some historical material may still use the old name.