Open Standard · v1.0.0 · AI-Driven Development

The standard
for AI-driven
development.

AI-DRIVEN DEVELOPMENT LIFECYCLE  ·  AUTHORED BY 10QBIT TECHNOLOGIES

"Humans define intent and validate outcomes. AI executes craft and maintains continuity. Every decision is preserved. Every output is auditable. Quality is a system property, not a human heroic act."
Read the Standard → View on GitHub Get Certified
Engineering Track

Seven phases.
One cadence loop.

A directed lifecycle that never terminates. Each cycle feeds the next through the Cadence Loop — the mechanism that makes AIDDLC a self-improving system, not a one-time process.

The Standard

Five principles.
No exceptions.

Implementations may extend these principles but may not contradict them. Any implementation that violates the Governing Principle is not AIDDLC-compliant.

01
Intent over implementation
Humans own the what and why. AI owns the how. The what and why are written before AI execution begins — they are the acceptance criteria, the user stories, the compliance requirements. AI may propose options, but selection belongs to a human.
Violation: allowing AI to define scope or generate acceptance criteria without human sign-off
02
Context never dies
Every decision feeds every future decision. There is no re-discovery, no repeated elicitation, no knowledge lost at handoffs. The Context Layer accumulates across all phases and is never overwritten — only extended.
Violation: beginning a phase without loading prior-phase context; repeating research already performed
03
Gates before generation
No AI generation begins without human-approved acceptance criteria. Phase gates are formal checkpoints. Task gates are lightweight checkpoints within phases. Both require human approval. AI may prompt for gate completion but may not self-certify gate passage.
Violation: generating output then deriving criteria from it; AI self-certifying gate passage
04
Continuous validation
Validation is not a phase. It is an activity present in every phase. The Validation phase is comprehensive integration validation — it does not substitute for in-phase validation. Each phase has defined validation activities that must complete before the gate.
Violation: deferring all validation to Phase 6; passing a phase gate without completing in-phase validation
05
Auditable by design
Every output, decision, and rationale produced during the lifecycle must be attributable, timestamped, and retrievable. This is the default operating mode of an AIDDLC implementation — not an audit-mode feature activated on request.
Violation: undocumented decisions; inability to trace a deployed feature to its originating requirement
Architecture

Two tracks.
One standard.

Engineering Track and Product Track operate in parallel, share the same governing principles, and exchange requirements and features at defined handoff points.

Engineering Track
How AI-assisted teams build software
Gate-controlled · Compliance-aware · Context-accumulating · 7 phases
FoundationIntelligence Brief · Compliance Matrix
Discovery & IntelligenceRequirements Canvas
Architecture & DesignADRs · Security Model
SpecificationAcceptance Criteria Set
BuildDecision Log · AI Contribution Record
ValidationCompliance Evidence Package
Deploy & LearnLearning Report · Cadence Loop
requirements
→
←
features
Product Track
How AI products find market fit
Model selection · AI-native metrics · Value realisation · 4 phases
DiscoverProblem-AI Fit · Model Selection
PrototypeLLM Benchmarking · UX Validation
DeployA/B Testing · AI-native Metrics
ScaleToken ROI · Model ROI
COMMUNITY DEVELOPMENT AREA
Product Track methodology is an active community contribution area. Domain experts are invited to develop the framework further.
Compliance

Built for regulated
environments.

AIDDLC artifacts map to universal regulatory concerns applicable across all frameworks and jurisdictions. Industry Profiles make it locally actionable — without making the core standard jurisdiction-specific.

Core Standard — Universal
AIDDLC Standard v1.0
Risk Register · Compliance Matrix · Audit Trail · Evidence Package · Sign-off Record
INDUSTRY PROFILES — maps the standard to specific regulators and jurisdictions
🇬🇧
UK Healthcare
MHRA · CQC · GPhC
NHS DTAC
v1.0 — Authored by 10QBIT
🇬🇧
UK Financial Services
FCA · PRA
Consumer Duty · AML
v1.0 — Authored by 10QBIT
🇺🇸
US Healthcare
HIPAA · FDA · ONC
21st Century Cures
Community Development Area
🇪🇺
European Union
GDPR · EU AI Act
MDR · ISO 27001
Community Development Area
Reference Implementation

Proven in production.

The AIDDLC Standard is not theoretical. It was developed through and validated by a live production deployment in a regulated healthcare environment.

Reference Implementation
Club Health OS
A regulated healthcare SaaS platform serving clinical weight management across 21+ clinics in the UK. Built end-to-end on the AIDDLC Standard — every phase, every gate, every artifact.
7
phases
completed
21+
clinics in
production
4
regulatory
frameworks
Completed phases
FoundationComplete
Discovery & IntelligenceComplete
Architecture & DesignComplete
SpecificationComplete
BuildComplete
ValidationComplete
Deploy & LearnActive
The reference portal used on Club Health OS is open-source and available at github.com/aiddlc/reference-portal

Ready to build
to the standard?

Read the specification, implement it, get certified. The founding certification cohort is open — limited to five organisations.

Read the Standard → Reference Portal Certification Programme