There is one question I now ask every leadership team I meet: who leads AI here?
The answers are revealing. Sometimes there's a pause. Sometimes it's "IT, sort of", or "we're working it out". Sometimes four people answer at once, each describing a different part of the picture. Very rarely does one person say "I do", and then describe a plan.
Large organisations have moved quickly on this. IBM's 2026 CEO study, a survey of 2,000 chief executives, found that 76% of their organisations now have a chief AI officer, up from 26% a year earlier. Most mid-sized businesses can't justify a full-time executive for AI, and so far most haven't found another way to fill the gap. AI is arriving anyway.
Everyone owns a piece
In most organisations, AI is already owned, just not as a whole.
| Who | What they own |
|---|---|
| IT | The tools: what is approved, licensed and supported |
| Security | The risk: data, access and threats |
| Finance | The cost: seats, subscriptions and the bill at the end of the month |
| The business | The ambition: where AI could make or save money |
| HR | The people: skills, roles and how staff feel about it |
Each does its part well. The trouble is the gaps between them, and the symptoms are familiar:
- pilots that work, but never scale
- a policy that exists, but isn't followed
- the same kind of tool bought twice, by two teams
- agents running in live systems with no one clearly accountable for them
- a board asking questions nobody can answer in one place
The people side shows the gap most clearly. In the same IBM study, CEOs said only 25% of their workforce uses AI regularly, even though 86% believe their people have the skills to. Nobody owns closing that gap, so it stays open.
Why IT can't simply own it
The natural answer is to hand AI to IT. It's where most technology lands, and IT leaders are usually the first to see the risks. But it rarely works on its own, for three reasons.
- IT owns the tools, not the outcome. Making AI an IT project turns it into a question of which products to approve, rather than what the business wants AI to achieve.
- The hard decisions aren't technical. Which roles change, which customer processes can run without a person checking, how much risk the business will accept: those are leadership decisions, and IT can't make them alone.
- Adoption is a people problem. In IBM's study, 83% of CEOs said AI success depends more on people's adoption than on the technology. That sits with line managers and the leadership team, not the service desk.
The same is true of security or finance. Each sees AI through its own lens, and each lens is right. None of them is the whole picture.
What leading AI actually involves
Leading AI isn't a job title. It's five responsibilities that someone has to hold together.
- Set the direction. A one-page plan: what AI should do for the business in three years, this year's goals, and this quarter's priorities.
- Decide in proportion to risk. A light touch for everyday tools, real scrutiny where customers, money or sensitive data are involved, and clarity about who approves what.
- Make sure everything has an owner. Every use of AI, and every agent, with a named person accountable for it before it goes live.
- Prove the value. Savings found and kept, and a regular view for the board of what AI is costing and returning.
- Hold the builders to account. Whether the work is done by your own teams or by partners, someone has to check it delivers what was promised.
None of these requires deep technical skill. All of them require someone senior enough to make decisions stick, and close enough to see what is actually happening.
Four ways to fill the gap
| Option | When it fits | The trade-off |
|---|---|---|
| A full-time chief AI officer | AI is central to the business, at real scale | Cost, and months to recruit for a scarce role |
| Add it to an existing leader (a CIO, CTO or COO) | AI use is modest, and that leader has time | Capacity, and a tendency to see AI through one function's lens |
| A fractional AI lead | AI is active and growing, but a full-time role isn't justified yet | Part-time, so it needs a committed sponsor inside the business |
| Combined AI and security leadership | A mid-sized business with neither a security leader nor an AI lead | One relationship wearing two hats, so the board must set the appetite for risk and high-risk decisions must sit above the lead |
The fractional route is how many mid-sized businesses already buy security and finance leadership. It is now becoming a normal way to buy AI leadership too.
Whichever route you choose, one thing stays inside the business. A named senior sponsor remains accountable for AI. Leadership can be fractional; accountability can't.
Three questions to ask this week
- If AI went wrong in a customer-facing process tomorrow, who would be accountable, and would they know?
- Is there one plan for AI, or several plans owned by different functions?
- Who tells the board what AI is costing and returning, and on what evidence?
If the answers are unclear, that's not a failing. It's where most organisations are, and it's the easiest gap to close before it becomes expensive. The first step is usually to name two people: a senior sponsor who is accountable for AI, and an AI lead who runs it day to day, often someone already in IT or operations. At Peruze, we help leadership teams agree those roles and what they are for, and where a business needs more capacity than it has in-house, a fractional AI lead from us can run the process for a set period, while the approvals stay with your own people.
Who leads AI in your organisation today, and how long did it take you to answer?
Sources
- IBM Institute for Business Value, 2026 CEO Study, announced 4 May 2026: a survey of 2,000 CEOs and equivalent leaders in 33 geographies, covering chief AI officers, workforce AI use and skills, and adoption as the driver of AI success