Living documentation that stays in sync with engineering, cloud, AI, and compliance activity. Documents update themselves as the organization changes.
UBC's AI Development Life Cycle Standard, covering intake, data, model & prompt engineering, evaluation, HITL, deployment, monitoring, and retirement.
Define a repeatable, auditable lifecycle for building and operating AI systems at UBC — voice agents, LLM applications, RAG systems, and classical ML models — with explicit guardrails for patient safety, PHI handling, and model risk.
All AI systems used by UBC internally or delivered to clients, including third-party foundation models called via API.
AI-DLC extends SDLC. Every AI project also follows SDLC governance where applicable (change control, secure coding, CI/CD). AI-DLC adds:
| # | Phase | Governing SOP |
|---|---|---|
| 1 | Use-Case Intake & Ethical Review | SOP-AIDLC-01 |
| 2 | Data Sourcing, Lineage & PHI Handling | SOP-AIDLC-02 |
| 3 | Dataset Curation, Labeling & Bias Assessment | SOP-AIDLC-03 |
| 4 | Model / Prompt Design & Guardrails | SOP-AIDLC-04 |
| 5 | Evaluation (accuracy, hallucination, safety, red-team) | SOP-AIDLC-05 |
| 6 | Human-in-the-Loop Review & Sign-off | SOP-AIDLC-06 |
| 7 | Deployment, Canary & Model Registry | SOP-AIDLC-07 |
| 8 | Monitoring, Drift & Feedback Loop | SOP-AIDLC-08 |
| 9 | Incident, Rollback & Retirement | SOP-AIDLC-09 |
| Tier | Description | Governance |
|---|---|---|
| Tier 1 | Patient-facing, autonomous decisions | Executive approval, quarterly review |
| Tier 2 | Assists human decisions on PHI | HITL required, monthly review |
| Tier 3 | Internal productivity, no PHI | Standard review |
Model cards, evaluation reports, HITL sign-offs, and drift dashboards are controlled documents.
Maps to NIST AI RMF, EU AI Act (limited-risk category), and ISO/IEC 42001:2023 AIMS.