AIGP Study Guide
Module 6: Governing AI Development · BoK III.A

Stakeholders - Who, What, and the Hard Calls

Engage stakeholders early, agree the goal, and decide who owns the failures. When values clash - e.g. more accuracy than privacy - the organisation must choose, get agreement and document that decision.

Engage early, agree the goal, and decide who owns the failures. Then let each function ask its own questions.

The cast. Common stakeholders → AI governance officers, privacy experts, security experts, sometimes procurement, always subject matter experts and legal.

  • Define and agree the goal of the AI → and whether AI even suits the mission and purpose
  • Establish parameters for success and a meeting frequency for continuous evaluation
  • Establish who is ultimately responsible for risks, mitigations and system failures after implementation
  • ID training data needed and applicable policies → sector-specific too, e.g., does HIPAA cover the training data in healthcare
  • What if the AI performs poorly → impacts on individuals and the organisation, and the risk tolerance per scenario
Competing values

There is no single perfect answer when values clash → e.g., a requirement for more accuracy than privacy. The organisation must decide which to prioritise, get the stakeholder group to agree, and document that decision. Then communicate identified risks and mitigations → to stakeholders, the organisation, partners and anyone receiving shared data results, tailored by audience.

Role-specific questions stakeholders ask
FunctionTheir questions
Legal & complianceWhat specific requirements will impact this system? What liability might this model create?
Marketing, procurement, salesWhat opportunities will this create? Which competitive advantage should be prioritised?
LeadershipIs the AI consistent with our values and business model - human rights, environmental impacts?