Who should own generative AI across the organisation?
Generative AI works best when ownership is shared but clearly assigned. Research suggests executive sponsorship, cross-functional governance and business-unit execution each matter. The practical question is not who “owns” AI alone, but how leaders coordinate risk, value and learning across the organisation.
Generative AI is moving fast enough to tempt organisations into a single-owner model: one executive, one program, one answer. The evidence points in a different direction. Stanford’s AI Index shows adoption spreading quickly, while responsible AI practices are lagging and incidents continue to rise 1. MIT research suggests organisations are already dealing with two distinct forms of generative AI: broad productivity tools and tailored solutions for specific business problems 3. That combination makes clear ownership essential, but not centralised ownership alone. The real task is to match decision rights to risk, use case and accountabilities across the organisation.
Start with a portfolio view, not a title
The first practical step is to stop asking who should “own AI” in the abstract. MIT Sloan’s research distinguishes between broadly applicable tools and tailored solutions, and those two categories need different controls, budgets and success measures 3. A chatbot used by many employees is not the same governance problem as a bespoke model embedded in a regulated workflow.
In practice, that means ownership should sit at more than one level. Executive sponsorship belongs with the senior leader accountable for transformation or strategy. Day-to-day stewardship of enterprise-wide use cases should sit with a cross-functional AI lead or programme owner. Individual functions should retain ownership of their own use cases, because they understand workflows, risks and service impacts best.
Give the centre coordination, not control
Harvard Business School’s research on leading AI teams suggests that effective leadership depends less on talking more and more on asking, listening and revising course when new information emerges 2. That translates well to AI governance. A central AI office should coordinate standards, not act as a bottleneck. Its role is to set guardrails, convene the right people, and keep decisions moving when legal, technology, risk and business priorities collide.
An effective centre can maintain approved tool lists, model-use principles, procurement criteria and review pathways. It can also stop duplication, where different parts of the organisation unknowingly buy overlapping tools or pursue inconsistent pilots. The limit of central control is that local teams still need enough freedom to test, learn and adapt. If every decision is escalated, adoption will stall and shadow AI is more likely to grow 3.
Make risk ownership explicit
One of the clearest findings in MIT’s work is that unsanctioned use of public tools can create data, intellectual property, privacy and security risks 3. Those risks are not evenly distributed. A marketing team experimenting with drafting support faces different exposures from a finance team using generative AI to summarise sensitive material. Ownership should therefore follow risk, not enthusiasm.
A useful approach is to assign named accountability for data protection, legal review, cyber security, people impacts and process quality. Business leaders should not assume technology teams will carry all of this. Nor should legal or risk teams be expected to drive adoption. The organisation needs a shared control model in which each function owns its part of the risk picture, and the executive sponsor owns the escalation path when trade-offs arise.
Treat adoption as a leadership and change issue
MIT Sloan’s discussion of generative AI leadership stresses the importance of trust, transparency and a clear narrative from senior leaders 4. That matters because employees will not sustain use of generative AI if they think it is being imposed without explanation. Ownership, then, must include responsibility for communication: why the organisation is using AI, what it will and will not be used for, and how people will be supported as tasks change.
This is also where practical leadership discipline matters. Harvard’s findings suggest that good AI-era leadership is active and curious rather than performative 2. Leaders should create regular forums where teams can raise problems, show what they are learning, and compare results across functions. From an Authenticate advisory perspective, the strongest AI ownership models are usually the ones that make learning visible without making every experiment political.
Use business units for value, not just experimentation
The organisation’s business units should own the value case. They are best placed to identify where generative AI can improve customer interaction, knowledge access, content production, case work or internal service delivery. MIT’s research on tailored solutions notes that value depends on choices such as whether an organisation buys, builds or enhances a solution 3. Those decisions cannot be made well from the centre alone.
To make this work, each unit should have a named business owner, a technical counterpart and a risk partner. Together they should define the problem, test the use case, and decide whether the solution is worth scaling. This avoids a common failure mode: pilots that look impressive in isolation but never connect to a business process, operating model or workforce plan.
A workable ownership model for executives
A sensible governance structure usually has three layers. At the top sits executive sponsorship, which sets intent and arbitrates trade-offs. In the middle sits a cross-functional AI steering group, which defines standards, approves priority use cases and manages organisational learning. At the edge sit business and functional owners, who run the work and take responsibility for local outcomes. That structure is flexible enough to handle broad tools and tailored applications differently, while still giving employees a clear place to go 3.
The practical test is simple: can the organisation answer who is accountable for tool approval, use-case prioritisation, data protection, workforce impact and escalation? If the answer is fuzzy, ownership is too diffuse. If every answer points back to one executive, ownership is probably too narrow. Generative AI needs a model where leadership, governance and execution each have defined jobs, with enough clarity to keep pace with change and enough trust to keep people engaged 14.
Sources
- The 2026 AI Index ReportStanford University — Human-Centered AI · University research · link checked 15 October 2026
- Think You’re a Good Leader? Try Leading an AI TeamHarvard Business School AI Institute · University research · link checked 15 October 2026
- How to manage two types of generative artificial intelligenceMIT Sloan School of Management · University research · link checked 15 October 2026
- Who should be the head of generative AI — and what they should doMIT Sloan School of Management · University research · link checked 15 October 2026
Further professional reading
External professional resources, separate from the numbered sources above. These are not university studies, and listing them does not indicate any collaboration, certification, affiliation or endorsement.
- Global Institute of Directors (opens in a new window)External professional resource · governance and directorship
- World Coaching Institute (opens in a new window)External professional resource · coaching practice