Who Should Own the AI Budget—and Why This Is a Terrible Question

Artificial intelligence has quickly moved from being an experimental technology to becoming one of the largest budget priorities in most organizations. In many companies today, a significant portion of discretionary spending is being redirected toward AI initiatives, pilots, tools, talent, and infrastructure. As usually happens in corporate environments, budget allocation is never just about money—it is also about influence. The function that owns the budget often shapes the agenda, sets priorities, and gains internal strategic visibility. Unsurprisingly, this has led many executives to ask what appears to be a perfectly reasonable question: Who should own the AI budget?

Should AI investments sit under IT because of the infrastructure and security implications? Should strategy own it, given AI’s potential to reshape business models and competitive positioning? Should innovation lead, as the team traditionally tasked with exploring emerging opportunities? 

Increasingly, some organizations are even creating dedicated AI leadership roles (Chief AI Officer) and separate AI functions to centralize responsibility.

While logical on the surface, this is ultimately the wrong question.

The assumption behind the question is that AI is a standalone capability that can be isolated, managed, and deployed as its own function. This is precisely where organizations risk making a familiar mistake. Anyone that’s been around corporate innovation in the past 10–15 years ago knows what I’m talking about.

Around a decade ago, companies had a nearly identical debate about innovation. Faced with pressure to “innovate or die,” many organizations responded by creating innovation departments, appointing Chief Innovation Officers, and building dedicated innovation fictions.

The intention was sound. The outcome, in many cases, was not. Leading to the raise of innovation theater and the fall of countless careers.

By separating innovation from the rest of the business, companies unintentionally turned it into a silo. Innovation teams often became disconnected from the operational realities, customer problems, and profit-and-loss responsibilities that define business priorities – all while giving them a false sense of superiority relative to their peers in the line business. Over time, innovation was increasingly perceived as something adjacent to the business rather than embedded within it. In many organizations, it gradually lost relevance—not because innovation ceased to matter, but because it became someone else’s responsibility.

There is a growing risk of repeating this same pattern with AI.

In an attempt to accelerate adoption and impose structure, organizations are establishing AI centers of excellence, ring-fencing AI budgets, and appointing executives to oversee AI strategy (Chief AI Officer). These moves are understandable through the lens of traditional management practices. However AI introduces unprecedented complexity around data governance, security, model management, compliance, and vendor selection. 

Some degree of coordination is not only useful, but necessary. However, centralization often comes with unintended consequences, a really significant dark side.

When AI is owned by a dedicated team or function, business units naturally begin to outsource responsibility. Rather than building internal capabilities, departments start waiting for the AI team to prioritize their requests. Demand quickly outpaces the capacity of a small centralized group, creating bottlenecks across the organization. 

At the same time, solutions risk being developed further away from the people who best understand the underlying business problems.

The result is a familiar organizational pattern: AI remains strategically important but operationally constrained.

This reveals the false dichotomy at the center of the debate. Organizations are often forced into choosing between full centralization and full decentralization, as if these were the only available models. Neither is likely to succeed (again we have seen this with innovation)!

So here is the conundrum: a fully centralized AI model may create alignment and governance, but it also introduces dependency and slows execution. A fully decentralized model may encourage experimentation and local ownership, but it often results in fragmented technology choices, duplicated efforts, inconsistent standards, and unmanaged risk.

The objective, therefore, should not be choosing one extreme over the other. The real design challenge is building an operating model that enables distributed ownership with centralized coordination (and governance).

This requires reframing the problem entirely.

The most important question is not who owns the AI budget, but how organizations ensure AI adoption happens at scale and as close as possible to where value is created. AI is not a departmental initiative; it is a general-purpose capability with applications across virtually every function. Marketing teams can use AI to improve customer segmentation and campaign performance. Operations teams can optimize workflows, forecasting, and supply chain efficiency. Finance teams can strengthen scenario planning, risk analysis, and decision support. HR teams can redesign talent acquisition, onboarding, and workforce planning.

The value of AI is realized through application, not ownership.

This is why every department should be responsible for identifying, prioritizing, and implementing AI opportunities relevant to its own domain. Business teams are far better positioned to understand where friction exists, where decisions can be augmented, and where automation can create measurable impact.

Yet distributed responsibility does not mean unmanaged responsibility.

One of the common failures in organizational transformation is assuming that simply mandating adoption is enough. From our experience it rarely is. 

We have seen this repeatedly in digital transformation, agile adoption, and innovation initiatives. Declaring AI a priority for everyone without building the underlying system for execution merely creates symbolic alignment.

For distributed AI adoption to work, organizations need several foundational elements in place. 

• First, teams require sufficient AI literacy and capability building. Without a baseline understanding of what AI can and cannot do, ownership becomes performative rather than practical. 
• Second, incentives must be aligned. If business leaders are not measured on AI adoption or operational improvement enabled by AI, other priorities will inevitably dominate. 
• Third, organizations need shared infrastructure, tooling, and access layers that reduce friction and prevent every function from building its own disconnected ecosystem. Finally, governance remains essential—but its role must evolve.

This is where many organizations misunderstand leadership roles such as Chief AI Officer. The issue is not the existence of these positions. In fact, many companies can benefit from strong AI leadership. The problem emerges when these roles become de facto owners of all AI execution. In that model, the organization inadvertently reinforces the idea that AI belongs to a specialist function.

A more effective interpretation of AI leadership is as an enabling layer rather than a controlling one. AI leaders should focus on building organizational capability, setting governance frameworks, defining standards, managing risk, and accelerating knowledge transfer across departments. Their success should not be measured by how much AI they directly control, but by how effectively they enable the rest of the business to adopt and scale it.

This distinction is subtle, but strategically important.

Ultimately, the question of AI budget ownership feels attractive because it offers a simple governance shortcut. It suggests that if the right function controls the budget, the organization will naturally get AI right. Experience suggests otherwise.

AI does not need to be owned as a separate corporate domain. It needs to be embedded into how the organization operates.

Which leads to a far more important question—one that organizations should arguably be asking instead:

How do we make AI everyone’s responsibility without making it no one’s job?

This is a more difficult question because it cannot be solved through org charts or budget lines alone. It forces leaders to think more deeply about operating models, incentives, governance, capability building, and organizational design.

But unlike the original question, it points in the right direction.

Because in the long run, organizations will not generate ROI from their AI projects by deciding who owns AI. They will generate ROI by ensuring AI is embedded where the problems actually are.