Discovery, deployment and adoption planning with clients
What the exam tests
This domain checks whether you can run discovery, scope a deployment, and plan adoption with a client-linking use cases, stakeholders, and success metrics to governance configuration.
Discovery and adoption rules
- Start with business objectives, AI use-case inventory, regulatory scope, current governance maturity, and risk appetite.
- Map stakeholders: executive sponsor, risk/compliance, model owners, data and AI teams, legal, audit, and end users.
- Define success metrics before configuration: review cycle time, inventory coverage, evidence completeness, risk-assessment completion, and monitoring coverage.
- Deploy in phases: pilot on a small set of representative use cases, configure taxonomy and workflows, integrate, test, train, then scale.
- Adoption needs a governance council, clear roles, champions, and communication; technology alone does not change behavior.
Worked example
An insurer has eight generative AI use cases. A 10-week pilot governs two: customer-service summarization and claims triage. Discovery finds no central inventory and inconsistent approvals. The pilot configures risk tiering, a fact sheet, an approval workflow, and monitoring for hallucination and PII leakage. Success metrics are 50% faster approvals and 100% evidence completeness for pilot models. After the pilot, the client extends to the remaining use cases and adopts a monthly governance council. The reasoning is to prove operating value on representative risk before scaling.
Traps that cost marks
- Starting with tool configuration before agreeing on use cases, risk appetite, and decision rights.
- Running a pilot without baseline metrics, then being unable to show value.
- Treating adoption as training only; role changes, incentives, and governance forums matter.