Chief AI Officer:
permanent role or transitional fix?

Spreadsheets changed the work, not the organisation chart

VisiCalc was released in 1979 to remove the drudgery of rebuilding financial models whenever an assumption changed. It helped turn the personal computer into a serious business tool and was followed by Lotus 1-2-3 and, eventually, Excel.¹

No widespread executive role called Chief Spreadsheet Officer followed it into the organisation. The capability migrated into finance, planning, operations and management. Spreadsheets became part of ordinary work, while specialist modelling and control remained attached to the functions that used them.

That history makes the Chief AI Officer worth questioning. If spreadsheets could transform analysis without acquiring permanent C-suite ownership, why should AI be different?

The comparison is useful, but it cannot settle the answer. Spreadsheets execute formulae defined by users. AI systems can generate text, classifications, predictions, recommendations and decisions; they may interact with data, software and people across several stages of work. Their usefulness and failure modes vary sharply by context. AI is not simply a more capable spreadsheet. It reaches further into judgement, authority, accountability and the design of work.

Why the role has appeared so quickly

The title is spreading because responsibility for AI is currently fragmented. One executive owns technology, another data, another cybersecurity, another people, another capital allocation, another operations, and another legal or regulatory risk. Business and service leaders own the outcomes, but may not understand the systems changing how those outcomes are produced.

IBM's 2026 CEO study offers a striking signal. Seventy-six per cent of 2,000 CEOs and equivalent senior leaders said their organisations had a Chief AI Officer, compared with 26 per cent in IBM's prior-year research.² The figures should not be read as a universal market census. The studies were commercially interested, self-reported and not a longitudinal panel of the same organisations. They do, however, show how rapidly formal AI ownership has entered executive discussion.

The CAIO is therefore not necessarily a vanity title or a sign that boards have mistaken fashion for organisation design. It can be a rational response to a genuine coordination problem: AI has become important before the surrounding responsibilities have found a stable home.

The responsibility is currently homeless

A useful CAIO sits across the gaps between established functions. The role may connect strategy to use-case selection, technology to workflow redesign, data to evaluation, risk to operating controls, workforce planning to adoption, and investment to measurable organisational value.

None of those responsibilities is entirely new. What is new is the need to align them at the speed and breadth with which AI is entering work. Without a coordinating mechanism, organisations tend to accumulate pilots, policies, licences and local workarounds without a shared portfolio, common evidence standard or clear route from experiment to dependable capability.

AI is not merely a more powerful spreadsheet

Spreadsheets certainly created errors, shadow models, audit problems and misplaced confidence. The distinction is not that one technology is safe and the other risky. It is that AI can operate across a less legible chain of data, models, prompts, tools, human review and downstream action.

The US National Institute of Standards and Technology describes AI risk governance as a cross-cutting activity that runs through the organisation and the system lifecycle. Its framework calls for clear roles and lines of communication throughout the organisation, while executive leadership remains responsible for decisions about AI risk.⁷

That combination matters. AI's breadth justifies stronger central coordination. Its dependence on context makes central ownership of every outcome impossible. A CAIO may establish the architecture for responsible use, but cannot know every customer promise, clinical exception, operational dependency, professional judgement or service consequence across the organisation.

The people who own the work must remain responsible for what the work becomes. The people who own data, security, legal risk, procurement, finance and workforce design must retain their responsibilities. AI leadership should connect those accountabilities, not replace them.

A title can solve an ownership gap - or hide one

A well-designed CAIO role creates a visible point of coordination. It can secure executive attention, arbitrate priorities, standardise evaluation, reduce duplicated effort, establish shared tooling and force difficult questions about value, risk and workforce consequences.

A badly designed role gives everyone else somewhere to send the problem. The CIO can say AI strategy belongs to the CAIO. Functional leaders can say governance belongs to the central team. Risk specialists can treat operational outcomes as a technology matter. The board can confuse appointment with capability.

IBM's own 2026 findings contain this tension. In the same survey that reported rapid CAIO adoption, 85 per cent of respondents said functional leaders must become technology experts in their own domains, while 79 per cent said they were decentralising decision-making and distributing accountability.² The signal is not that one officer will own AI. It is that organisations are trying to centralise coherence while distributing responsibility.

IBM itself provides a commercially interested but useful counter-example. Its account says the company has no formal CAIO. A senior transformation and operations leader coordinates enterprise change, while individual leaders remain accountable for AI adoption in their teams.³ That is not proof of a superior model. It does show that a CAIO-shaped function can exist without a standalone title.

The Chief Digital Officer precedent

The closest organisational precedent is not the spreadsheet manager. It is the Chief Digital Officer.

A 2018 Oxford Saïd report, based on interviews with 41 Chief Digital Officers across 13 countries, described a role created to drive far-reaching change but expected by some holders to become obsolete once digital became the normal operating model.⁴ A larger archival study of US public companies found that Chief Digital Officer roles emerged as a distinct structural choice and varied substantially between generalist and domain-specialist forms.⁵

The two findings belong together. Transformation roles can be genuinely necessary without having one permanent design. Some disappear, some merge into established posts, and some persist because the organisation retains a durable specialist need.

AI may follow the same path, but not mechanically. Digital became an adjective that eventually lost much of its separating power. AI may leave a larger permanent residue of model engineering, evaluation, assurance, data infrastructure and high-impact governance. The question is therefore not whether the function disappears. It is which parts remain central and which parts become normal managerial responsibility.

Centralise the transition; distribute the capability

The strongest CAIO mandate treats the role as the architect of an ownership system, not the owner of AI.

Centralise what benefits from coherence: enterprise purpose, portfolio priorities, common platforms, evaluation methods, minimum controls, vendor principles, shared learning and the escalation of genuinely cross-functional decisions.

Distribute what depends on context: use-case selection, workflow design, service quality, human oversight, employee development, value capture and accountability for the result. Those decisions should sit with the leaders who understand the work and carry its consequences.

Specialist functions should retain their own authority. Data leaders govern data. Security leaders govern security. Legal and risk teams interpret obligations and risk appetite. Finance tests investment and value. HR and line management shape skills, roles and incentives. Procurement controls the commercial boundary. The CAIO coordinates the system in which those responsibilities meet.

US federal policy illustrates this distinction unusually clearly. OMB Memorandum M-25-21 requires agencies to retain or designate a CAIO, but permits the designation to sit with an existing CIO, CDO, CTO or similar official with sufficient expertise and authority. The same memorandum says senior leaders must distribute responsibilities and accountability across the agency.⁶ The function is mandatory in that setting; a new standalone executive seat is not.

Permanent responsibility, conditional function, potentially temporary title

Three separate decisions are often collapsed into one.

The responsibility is permanent. Organisations will continue to need decisions about AI strategy, evidence, risk, work design, people, suppliers, value and learning.

The central function is conditional. A dedicated enterprise AI office may remain valuable where scale, common infrastructure, high-impact use, research intensity or cross-functional portfolio choices justify it.

The title is reviewable. The CAIO may remain, merge with technology, data, strategy or operations, or become an additional designation carried by an existing leader.

A permanent CAIO is most defensible when AI is central to the organisation's product, mission or competitive position; when a large shared model and data portfolio requires enterprise authority; when high-impact uses need continuing specialist assurance; or when the CEO needs a durable horizontal executive to arbitrate decisions that no existing function can credibly own.

A transitional or combined role is more plausible when the main problem is adoption coordination, when shared standards and platforms can be established, and when existing leaders can progressively absorb responsibility for AI within their own functions.

Smaller or less complex organisations may need neither a new office nor a new C-suite title. They still need a named senior owner, clear decision rights and a cross-functional mechanism proportionate to the risk.

What a sensible board should decide now

The first board question should not be, "Do we need a Chief AI Officer?" It should be, "Which AI decisions are currently unowned, fragmented or trapped between functions?"

From there, the organisation can decide whether the answer is a new executive, a combined designation, a transformation leader with an AI mandate, or a cross-functional council supported by existing executives. The structure should follow the missing decisions, not the fashion of the title.

Where a CAIO is appointed, the mandate should contain transfer conditions from the beginning. Which capabilities must business and service leaders acquire? Which standards should become routine? Which controls belong in existing functions? Which central specialist capabilities are likely to remain? When will the board reconsider the shape of the role?

This does not weaken the CAIO. It gives the role a more serious purpose. A transformation executive without a defined end state is easily trapped between permanent evangelism and permanent firefighting.

The title is a means of creating ownership, not a substitute for it

The Chief AI Officer may prove to be one of the defining executive roles of the next decade. In some organisations it will deserve a permanent place. In others it will be the temporary structure needed to build strategy, standards, shared capability and executive confidence before responsibility settles into the wider operating model.

Either outcome can be successful. The failure is to create the title without designing how ownership will work around it.

Spreadsheets became transformative because they stopped belonging to a specialist and became part of how managers worked. AI will not follow exactly the same path, but it is likely to follow the same organisational direction: central expertise where it adds leverage, distributed responsibility where context and consequence reside.

A successful Chief AI Officer may ultimately be judged by how thoroughly they make AI part of everybody else's job, without allowing accountability to belong to nobody.