AI LEADERSHIPFractional chief AI officer: who owns AI when nobody has the time?

What does a chief AI officer do?

The role is new enough that the clearest job descriptions come from governments, which had to write them down.

The US Office of Management and Budget’s April 2025 memo required every federal agency to name a chief AI officer within 60 days. Their duties include promoting AI adoption through a governance and oversight process, serving as the senior adviser on AI to the agency head, keeping an inventory of AI use cases, putting processes in place for high-impact AI, advising on an AI-ready workforce and guiding AI investment. [2]

Australia followed. The APS AI Plan, launched on 12 November 2025, gives every public servant “leadership and support from Chief AI Officers to promote adoption”. [3] Agencies had until July 2026 to appoint one of their existing senior leaders as chief AI officer, separate where possible from the AI accountable official who owns governance. In the words of the guidance, chief AI officers “lead the transformation”; accountable officials “enable it”. [4]

For a business, the job comes down to five things:

• Direction. Which problems AI should solve, in what order, tied to cost and growth targets.
• A register. Where AI is already used, by whom, with what data. Australia’s Guidance for AI Adoption starts with exactly this: decide who is accountable, and keep an AI register. [5]
• Rules people can follow. An AI policy, data rules and a risk process proportionate to the use case.
• People. Training for leaders and teams, so tools get used well, not just bought.
• Proof. A baseline before every pilot and a report on what it returned.

Companies are moving quickly. In IBM’s 2026 CEO study of 2,000 CEOs across 33 geographies, 76% of organisations had a chief AI officer, up from 26% in 2025. [1]

When does a mid-size company need one?

Not when the board asks for an AI strategy. When these start to happen:

• Several teams are paying for AI tools and nobody can list them.
• Staff are pasting customer or financial data into public chat tools and there is no policy.
• Pilots keep starting, and none has a baseline, an owner or an end date.
• A customer, auditor or regulator asks how you use AI and nobody can answer in one page.
• The board wants a number, and you have anecdotes.

The last one is common. McKinsey’s 2026 survey found only 37% of respondents attribute any EBIT impact to AI. [6] Gartner found only 13% of CIOs say their enterprise has realised significant or transformational value from AI, and puts the odds of a use case succeeding at one in five. [7] Somebody has to own the gap between using AI and earning from it.

Fractional, full-time or consultancy: which fits?

Three honest observations:

• Full-time makes sense later. When AI touches most of the business and you are past the first wins, a dedicated executive pays for themselves. Before that, it is a large cost ahead of proven value.
• Fractional needs an internal partner. A part-time leader brings judgement. Someone inside still has to chase the follow-through between visits.
• A consultancy is not a chief AI officer. It can give you a plan, train your people and build the systems. On its own, it does not sit in your leadership meetings or carry the agenda from quarter to quarter.

For most mid-size companies, the pattern we recommend: name an internal owner, bring in a fractional chief AI officer for senior judgement and governance, and revisit the full-time question in a year.

What should the first 90 days look like?

Whoever owns AI, fractional, full-time or an existing executive, this is a plan they can run:

Days 1 to 30: see what you have

• Build the AI register: every tool, every team, what data it touches.
• Publish a one-page AI policy staff can follow tomorrow.
• Interview the heads of finance, operations, sales and HR: where does work wait, and what would 20% better be worth?

Days 31 to 60: choose and start

• Score the use cases on value, feasibility, risk and how easily the result can be proven.
• Pick one pilot. Set the baseline, the owner and the stopping rule before anything is built. Gartner found organisations with rules for when a pilot should not reach production are 2.3 times more likely to be among the leaders. [7]
• Start training: leaders first, then the team that runs the pilot.

Days 61 to 90: prove and report

• Run the pilot against the baseline, with people reviewing every output.
• Report to the CEO and board: what was spent, what changed, what is next.
• Decide: widen, fix or stop. Then pick the second use case.

Vishal Kumar as your fractional chief AI officer

Quantana offers a fractional chief AI officer service, led by me. I serve as a fractional chief AI officer and adviser to several companies, from enterprises to startups. I am a Claude expert who advises enterprises and startups on getting real work out of Claude, ChatGPT and Copilot. I have taught AI strategy to business leaders at London Business School and trained 10,000+ professionals in AI, design thinking and storytelling. More about me.

What I do

• AI strategy and roadmap: which problems AI should solve first, tied to your cost and growth targets.
• Board and executive briefings: what AI can and cannot do for your business, in plain English, with numbers.
• Vendor and model choices: which tools to buy, which to build and which to skip. As of 10 November 2026, I recommend Claude as the best all-round AI tool for business teams.
• Governance and risk: the AI register, a policy people can follow and data rules, so a leak does not end the programme.
• Cost control and usage limits: the right mix of AI seats for people and API-based agents for repeat work, with usage caps for each person. At one manufacturing company, QSAP delivered 60% savings.
• Training so seats get used: paid seats nobody opens are the most common waste I see. In the teams we train, unused Claude seats go from about half to near zero.
• Pilots to production: a baseline and a stopping rule for every pilot, and oversight until the result is live and measured.

How an engagement works

1. Start with the roadmap. AI Compass gives us a readiness score, scored use cases, a phased plan with decision gates and what to do first, in 14 days. One fixed fee, never per user: From A$15,000 + GST.
2. Then a fixed monthly advisory retainer. I join your leadership rhythm, own the AI agenda with your executives and report progress to the CEO and board. One fixed fee a month, never per user.
3. Quantana delivers when you need it. AI Academy trains leaders and teams, and AI Labs builds and runs the systems on your data. See the AI transformation journey.

You keep the decisions. I bring the method, the judgement and the engineers behind me. For the method itself, read our guide to building an enterprise AI roadmap.

What mistakes should you avoid?

• A title without authority. If the owner cannot change how finance or operations work, the role becomes an IT side project.
• A strategy without a number. ChatGPT isn’t your AI strategy. Every item on the roadmap needs a baseline and an owner.
• Governance last. A one-page policy in month one prevents the data leak that ends the programme in month six.
• Too many pilots at once. Pick one, prove it, then widen.

General information, not legal advice. As at 10 November 2026.