Certified AI Project Management Professional

What you will be able to do

• Structure an AI project in stages with go and no-go gates tied to evidence
• Estimate work whose outcome depends on experiments and data quality
• Build a risk register that covers model, data and ethical risk
• Run hybrid delivery that combines plan-driven and iterative work
• Govern vendors and foundation-model services
• Hand over a live model with monitoring, ownership and retraining in place

Modules

1. 1How AI projects differ

   Uncertainty sits in the middle of the project, not at the start. The plan must allow for the model not reaching the target.

   • The AI project lifecycle from discovery to operation
   • Experiment-driven scope
   • Why 'done' includes monitoring
   • Typical failure points and their early signs

   Exercise. Map a past project of yours onto the AI lifecycle and mark where an evidence gate would have changed a decision.

2. 2Staged planning and evidence gates

   Breaking work into feasibility, proof of value, pilot and scale, each with a pass condition agreed with the sponsor in advance.

   • Defining pass conditions for each stage
   • Time-boxing experiments
   • Kill criteria and how to present them
   • Benefits tracking from the first stage

   Exercise. Write the gate criteria for moving a chatbot from proof of value to pilot.

3. 3Estimating under uncertainty

   Ranges, reference classes and option budgets in place of single-point estimates.

   • Three-point and range estimates
   • Reference-class forecasting
   • Budgeting for data preparation and labeling
   • Compute and inference cost estimates

   Exercise. Produce a range estimate for a document-extraction project and state the two assumptions that move it most.

4. 4Risk, quality and ethics

   A risk register that covers data, model performance, drift, security, regulation and reputation, with owners and triggers.

   • Model and data risk categories
   • Quality planning with evaluation sets
   • Regulatory awareness and impact assessments
   • Incident response for model failure

   Exercise. Add five AI-specific risks with triggers and responses to a project risk register.

5. 5Teams, vendors and foundation models

   Who you need, what to buy and how to contract for services whose behavior can change without notice.

   • Roles: data engineer, ML engineer, domain reviewer, product owner
   • Build, buy or adapt decisions
   • Contract terms for model services: data use, versioning, exit
   • Managing hybrid agile and waterfall teams

   Exercise. Draft five questions to put to a vendor of a hosted language model before contract.

6. 6Transition to operations

   A model in production needs an owner, monitoring, a retraining plan and a budget. The project is not closed until these exist.

   • Operational acceptance and runbooks
   • Monitoring for drift and performance
   • Ownership and RACI after handover
   • Lessons learned and benefits realization

   Exercise. Write the handover checklist you would require before closing an AI project.