Managing AI investment and accountability

When an organisation the size of Deloitte UK creates a brand-new Chief AI Officer role as part of a series of “strategic appointments”, it is worth asking what problem that appointment is actually solving. The answer, reading beyond the announcement, is likely an operational one. Someone needs to be accountable for how AI is governed, and what it costs. Yet many organisations have not yet worked out who that person is.

 

AI has crossed a threshold in enterprise adoption, with Generative AI now in use across virtually all large organisations. 45% are deploying it extensively. At that scale, this type of infrastructure needs clear ownership. AI gets embedded into existing applications and picked up by teams across the business for purposes that were not part of the original proposal. New use cases develop quickly, and ownership does not reset each time one does.

 

This results in AI spend that becomes genuinely difficult to track, cutting across technology budgets and application costs, and those areas already sit under different leaders with different mandates. The truth is, it’s often a structural problem, and it is the same problem cloud created a decade ago, which we’ve seen does not resolve itself.

 

 

AI accountability is moving up the agenda

Boards and executive teams are increasingly asking for clarity on AI spend. What does it cost once it is running? How is it performing against the goals that justified the investment? Where does the risk sit?

 

These are reasonable questions. For many organisations, AI has moved well beyond the experimental phase and into core operational spending. The expectation from leadership is shifting accordingly. It is no longer enough to report that AI initiatives are underway. There is now increased pressure to show what the investment looks like once it is live, and what the business is getting in return.

 

Many organisations are not yet structured to answer that with confidence. According to Flexera’s recent State of the Cloud report, only 47% of large enterprises have a defined AI governance lead. That figure reflects how early the structural response still is, even as the technology moves fast.

 

 

Spending without an owner

This is not a new problem. Cloud went through exactly this transition, and the industry found an answer in FinOps. It gave organisations a discipline specifically designed to own cloud cost, performance and accountability in one place. It became standard practice for organisations because unmanaged cloud spend had a habit of compounding quickly.

 

AI now shows a similar pattern. 53% of organisations cite security and compliance as the biggest barrier to scaling AI. That is usually not a technical limitation. It is a governance gap, and it is one that boards are increasingly being asked to account for without always having the structures in place to do so.

 

AI costs are also harder to see than cloud ever was. Unlike cloud, they do not arrive as a single consolidated bill. Nor are the AI costs the same month to month as adoption continuously spreads across almost every department. AI spending is embedded inside existing cloud spend, software contracts and tools that teams are already using.

 

By the time the full picture becomes visible, it is often larger than anyone anticipated.

 

Finance can see the spend but may not have the context to challenge it. Technology teams understand the architecture but are focused on delivery. The gap between those two positions is where cost and risk accumulate. FinOps closed that gap for cloud. The same discipline, applied to AI, can do the same.

 

 

Closing the gap

Deloitte’s appointment is an early signal of where the broader industry will be heading. The organisations moving fastest on this are not waiting for the role to become standard. They are already identifying someone who can own the technicalities of AI spend end to end, whilst being able to engage credibly with board-level scrutiny.

 

That person needs to understand how AI costs are structured and where governance risk sits and be able to translate both into language that leadership can act on. That combination of skills exists. The FinOps community has been developing it for cloud for years. The Linux Foundation is following suit with the intent to build the Tokenomics Foundation.

 

AI investment will be held to the same standard as every other major cost centre. The organisations that put someone in charge of meeting that standard earliest will be better placed to govern and scale with confidence. That starts with deciding who owns it.

 

---

 

Marlon Oliver is SVP EMEA at Flexera

 

Main image courtesy of iStockPhoto.com and Natee127