| 🏠 Back to Exam Syllabus | 📺 RooCloud on YouTube | 🌐 RooCloud Practice Exams |
The AI Life Cycle: All 7 Stages from Plan to Retire
This fourth episode of the ISACA Advanced in AI Audit (AAIA) exam prep series walks through every phase an AI system passes through, from the first concept to the final shutdown. Along the way it explains the concept of latent risk, why different actors view the same system differently, and the controls that keep an AI deployment trustworthy across its full life cycle.
What this episode covers
- Latent risk in AI systems and how different actors perceive the same risk differently.
- Plan and Design — establishing use cases, stakeholders, governance, and prototypes upfront.
- Collect and Process Data — consent, cleaning, completeness, and dataset documentation.
- Build or Adapt Models — calibration, training, explainability, and Human in the Loop oversight.
- Test, Evaluate, Verify, and Validate — the difference between verification and validation, plus the five testing types.
- Deploy — piloting, legacy compatibility, technical documentation, and change management.
- Operate and Monitor — running a Quality Management System for continuous compliance.
- Retire or Decommission — formal sunset planning when risk exceeds appetite or the tool becomes obsolete.
Watch the full episode above for the worked examples and detailed explanations of each concept.
Frequently Asked Questions
What are the seven stages of the AI life cycle?
The seven phases are Plan and Design, Collect and Process Data, Build or Adapt Models, Test Evaluate Verify and Validate, Make Available for Use or Deploy, Operate and Monitor, and Retire or Decommission. Risk weaves through every single stage, and different actors may perceive those risks very differently.
What is latent risk in an AI system?
Latent means hidden or dormant. A system might look perfectly safe during early development, but as it encounters new information in the real world it adapts and new dangers can suddenly wake up. This is unique to AI and is why everyone involved shares responsibility for keeping the system trustworthy and fit for purpose.
What is the difference between verification and validation?
Verification asks whether you built the system correctly according to your engineering blueprints. Validation asks whether you built the right system for your actual business problem. You could perfectly verify a high-speed train engine, but if your business needed an airplane it fails validation.
What are the five types of testing in the AI life cycle?
The five testing types are model testing for performance metrics like accuracy and precision, stress testing with extreme inputs, comparative analysis against older baselines, bias and fairness checks to ensure equitable treatment of all demographics, and scenario analysis using hypothetical real-world situations to predict how the system will react.
📚 Master the ISACA AAIA Exam!
Ready to test your knowledge? Access chapter-specific Multiple Choice Questions (MCQs) and full-length practice exams for the ISACA AAIA certification at RooCloud.com. Solve the chapter-wise questions to reinforce this lesson before moving to the next episode.
Reference: This article is based on concepts discussed in The AI Life Cycle: All 7 Stages from Plan to Retire.