5.3.2. Cross-Functional AI Teams and Accountability
💡 First Principle: An AI initiative touches the business process, the technology, the data, the law, and the people affected, so the team that delivers it must include all of those perspectives from the start. And because shared responsibility tends to become no responsibility, every initiative needs one clearly accountable owner, with defined roles for everyone else.
A typical cross-functional AI initiative team includes:
- Business lead / product owner: accountable for the problem definition, outcomes, KPIs, and adoption; usually the single accountable owner.
- Technical experts: data scientists, ML and software engineers, data engineers, architects.
- Data owner or steward: ensures data access, quality, and permitted use.
- Legal and compliance: regulatory requirements, contracts, IP, data protection.
- Risk and security: risk classification, security controls, responsible AI requirements.
- Change and HR partners: communication, training, role changes, employee relations.
- End-user representatives: the people whose work changes, providing requirements and early feedback.
Accountability standards make the team work:
- A single accountable owner for each initiative, usually from the business, with decision authority matched to the initiative's risk tier.
- Defined roles (for example, a RACI, as in 4.2.1) so that every decision has an owner.
- Clear success criteria agreed at the start (KPIs, baselines, responsible AI requirements).
- Regular reviews at stage gates with documented decisions.
- Escalation paths to the governance council for conflicts or high-risk decisions.
Bringing legal, compliance, and risk in at the start (rather than at a final review) is the people side of governance by design (4.1.3). It also speeds delivery, because requirements are known early.
⚠️ Exam Trap: When asked how to structure an AI initiative team, prefer options that combine diverse expertise (business, technical, legal, compliance, risk, end users) with one clearly accountable owner and defined roles. Purely technical teams, or committees where everyone shares accountability equally, are distractors.
Reflection Question: A data science team built an AI pricing tool that sales managers refuse to use and legal has now flagged for review. What was missing from the team structure, and how would you restructure accountability for the next phase?
