Build stage

Stage 2 – Build

Stage 2 of the AI Lifecycle is the Build phase, where the team commits to a specific use case and works toward production deployment. This stage typically lasts 4–16 weeks and focuses on selecting and validating vendors, configuring the solution, and preparing for a controlled pilot.

Key Activities

• Vendor scoring against Vendor Index
• RFP issuance (VEN-02)
  • Functional fit to the chosen use case
  • Data, security, and privacy controls
  • Integration and deployment model
  • Support, SLAs, and pricing
• POC validation (VEN-03)
  • Technical feasibility and performance
  • Alignment with non-functional requirements (latency, reliability, observability)
  • Fit with internal tooling, data sources, and workflows
• Contract negotiation
  • Pricing and usage tiers
  • Data ownership, retention, and model training rights
  • Security, compliance, and audit rights
  • SLAs, support, and exit clauses
• Configuration
  • Environment setup (dev/test/pre-prod)
  • Model and feature configuration
  • Integrations with identity, data, and application layers
• Pilot design (USE-02)
  • Target user groups and scope
  • Success metrics and KPIs
  • Guardrails, monitoring, and feedback loops
  • Rollout plan and change management approach
• Training (TAL-01)
  • End-user enablement and usage guidelines
  • Risk, compliance, and escalation procedures
  • Support playbooks and FAQs

Governance – Build-Stage Gate (GOV-15)

At the GOV-15 Governance Operating Cadence checkpoint, the following artefacts are required:

• Vendor selection rationale
• AI Bill of Materials (BoM) entry
• Risk Register draft entry
• Resourcing plan

Stage Outcomes

The Build stage concludes with one of three governance decisions:

• Deploy authorisation – proceed to deployment and productionisation.
• Hold-for-amendment – pause to address gaps (e.g., risk, performance, or resourcing issues).
• Rejection – stop the initiative or re-scope to a different use case or vendor.