When organizations begin AI governance work, they typically think about what expertise they're missing. They don't have regulatory expertise, so they hire a lawyer or a compliance officer. They don't have technical expertise, so they hire a data scientist or CTO. They don't have governance expertise, so they hire a Chief AI Officer. All of these people are useful. None of them is the right first hire.
The right first hire is someone who has successfully managed complex, cross-functional programs with multiple stakeholders and fifty-plus deliverables. A programme manager. Not because they need to write governance framework documents or understand the regulation, though those are helpful. But because AI governance is fundamentally a coordination problem, not an expertise problem.
Why AI Governance is a Coordination Problem
AI governance touches seven functions: Legal, Technology, Product, Operations, HR, Risk, and the Board. Each function owns part of the work. Legal needs to understand regulatory requirements and review contracts. Technology needs to own system architecture and implement controls. Product needs to make decisions about which systems are high-risk. Operations needs to own the AI system inventory and day-to-day oversight. HR needs to define roles and responsibilities for human oversight. Risk needs to own monitoring and incident response. The Board needs to understand the program and be assured it's on track.
No single person owns all of this. No single function owns all of this. The work requires coordination across seven functions with different priorities, different languages, and different definitions of success. A lawyer cares about compliance. A technologist cares about implementation. A product manager cares about user experience. A Board member cares about risk. Nobody naturally cares about integration and coordination.
That's the programme manager's job. They don't need to know more about regulation than anyone else. They need to make sure the lawyer and the technologist are talking. They don't need to know more about technology than anyone else. They need to make sure the technologist's implementation meets the lawyer's requirements. They own the master plan, the timeline, the deliverables, the dependencies, and the escalation paths.
What a Good Programme Manager Does
A good AI governance programme manager does five things well. First, they build the master plan. They understand all the work that needs to happen across all seven functions. They sequence it. They identify dependencies. They set realistic timelines. They identify what needs to happen in parallel and what needs to happen sequentially. They understand that the AI system inventory (which takes weeks) needs to be done before risk classification (which depends on knowing what you have).
Second, they set up governance cadence. They establish weekly stand-ups with team leads from each function. They establish monthly steering committee meetings with senior leaders. They establish quarterly board updates. The cadence creates accountability and visibility. Progress is tracked. Risks are surfaced. Decisions are made.
Third, they manage fifty-plus deliverables. They maintain a comprehensive list of what needs to be delivered: policies, procedures, technical controls, audit plans, training materials, documentation, roles and responsibilities, oversight processes, escalation procedures, and more. Each deliverable has an owner, a due date, and a status. When something slips, they escalate. When something is unclear, they clarify. They keep the program on track.
Fourth, they translate between stakeholders. The lawyer talks about regulatory requirements and liability. The technologist talks about infrastructure and data architecture. The product manager talks about user experience. The Risk leader talks about monitoring and controls. These are different languages. The programme manager speaks all of them. They translate regulatory requirements into technical requirements. They translate technical constraints into product trade-offs. They translate product changes into regulatory implications.
Fifth, they manage scope and escalation. Scope creeps in complex programs. Someone asks "shouldn't we also address AI procurement?" Someone else says "don't we need to think about AI in HR systems?" The programme manager understands what's in scope and what's out of scope. They manage what gets added to the program. When trade-offs are needed, they escalate to leadership rather than deciding themselves.
Why Expertise Specialists Struggle Without This
I've watched governance specialists start programs without a programme manager. They usually founder because they lack coordination. The specialist defines governance requirements, but they don't build a timeline that the organization can commit to. They don't sequence work across functions. They don't set up governance cadence. They don't manage fifty deliverables. Weeks go by with no visible progress. Stakeholders become frustrated. The program loses momentum.
I've watched lawyers start programs. They understand requirements better than anyone. But they don't understand implementation sequencing. They don't translate requirements into actionable work for the technology team. They set timelines that are unrealistic because they don't understand technical complexity. Six months in, they're frustrated because technical work isn't done, and the technical team is frustrated because legal keeps changing requirements.
I've watched technologists start programs. They understand what can be built and how long it takes. But they don't understand regulatory requirements. They don't understand how to translate technical controls into governance controls. They build tools that meet technical requirements but don't meet governance requirements. They're frustrated because the regulation seems to ask for things that aren't technically feasible.
A programme manager without deep expertise in any of these areas, but with the ability to coordinate across all of them, succeeds where specialists alone fail.
How to Evaluate Programme Manager Candidates
When hiring a programme manager for AI governance, look for three things. First, experience managing complex, cross-functional programs. Have they managed programs with more than four functions? Have they managed programs with more than fifty deliverables? Have they managed programs that required coordination with legal, technology, and operations? The deeper their experience with complex programs, the better prepared they'll be.
Second, ability to learn quickly about the domain. They don't need to be an AI expert or a regulation expert. But they need to be able to understand the key concepts, ask good questions, and connect dots between different areas. Ask them how they learned about a complex domain in a previous role. Look for intellectual curiosity.
Third, credibility to manage up. They'll be coordinating with senior leaders from multiple functions. They need to be taken seriously. They need the ability to push back on unrealistic timelines without being defensive. They need the ability to escalate conflicts and make them visible to leadership. This often comes from experience in roles where they had to manage across seniority levels.
The Strategic Outcome
Organizations with strong programme managers deliver on their governance programs. They hit their timelines. They manage scope. They translate requirements into implementation. They surface risks early. They build accountability and momentum. Organizations without programme managers get stuck in coordination failures. Expertise specialists work in silos. Requirements don't translate into implementation. Timelines slip. Momentum is lost.
The AI Act compliance program is a multi-year, cross-functional transformation. It requires coordination as much as it requires expertise. The programme manager is the person who makes that coordination happen. Hire them first. The specialists can follow.