A business can have a strong engineering team and still have nobody who owns its AI decisions. Those decisions include which models and vendors to use, what to build and what to buy, who to hire, and which pilots to stop. When the board asks for an AI plan, the decisions tend to land on a CTO who is already running something large.
A fractional Head of AI is one answer: a senior AI leader who works part-time beside the existing technical team and owns the AI calls. It is one of four routes. Choosing between them comes down to two questions. Who has to hold authority over the AI decisions? And how many days a week do those decisions need an owner? This piece covers where the role fits, what it covers, and how to tell when a different route suits you better.
What goes wrong when nobody owns the AI decisions
Without someone who owns the AI decisions, a capable engineering team can drift. Features ship one at a time with nothing joining them up. Pilots multiply without a business result anyone agreed to measure. Questions of regulation and risk come up late, once someone outside the team asks them.
The first question is who decides
Conversations about AI leadership tend to open with a job title. The title is best settled last, once you know who holds authority over the AI decisions and how many days those decisions need.
AI consultant vs fractional Head of AI
The difference between an AI consultant and a fractional Head of AI is authority. A consultant assesses and recommends, and someone in the business decides. A fractional Head of AI decides, part-time, and answers for the decision. Agathon's Fractional AI CTO page puts it in one sentence: "A fractional CTO holds decision-making authority, while a technical advisor makes recommendations for others to decide on."
A fractional leader therefore:
- Takes direct accountability for outcomes
- Works within the organisation's structure rather than alongside it
- Makes decisions rather than just providing recommendations
- Transfers knowledge continuously rather than at project boundaries
For many businesses the advisory arrangement is the right one. The task is to work out whether you need someone to improve your decisions or someone to take them.
Counting the days the decisions need
Authority settles what kind of help to get. Days settle how much. Choosing a vendor might take a few concentrated days and then a review each quarter. An AI team working on a live product raises questions every morning, such as whether a model change goes out or whether a contractor's estimate is credible. Estimate the days per month the decisions need over the next six months, and the route usually becomes clear.
Four routes, and when each one fits
Stretch the CTO you have, with independent advice
If the CTO already holds the authority and has the board's trust, moving AI decisions elsewhere creates a second centre of power over the same technology. In this route the CTO owns the AI decisions and an outside expert sharpens them, by comparing vendors, testing claims and pressure-testing a plan before it reaches the board. Agathon's AI leadership advisory page states the boundary: "AI leadership advisory is an external advisory role, while a Head of AI is an employee who runs the AI work day to day and is measured on it." The decisions, and the accountability for them, stay with the team.
The route is failing when AI decisions queue for a CTO who has no time, or when they get made by whoever is building. In both cases the business is missing an owner, and more advice will not supply one.
Take a hypothetical 60-person software firm whose CTO is halfway through a platform migration while the board asks for an AI plan by next quarter. If the migration ends within a few months and the CTO wants the AI work, advice that prepares the options fits. If it has a year left to run, the AI decisions will wait or drift, and one of the next routes fits better.
A fractional AI CTO, when there is no senior technical leader
Some businesses have no CTO to stretch. Agathon's service page describes this shape: "I own all technology decisions, with AI at the centre. The right shape when you have no senior technical leadership and AI is core to the product or the plan." The role owns architecture, vendors, technical hiring and the roadmap, and answers to the CEO and the board. For the role in detail, see what a fractional AI CTO decides.
Agathon's terms, as its pages state them: a Fractional AI CTO typically works one to three days a week, with a three-month minimum term and the days capped in writing. If the decisions need more days than the cap allows, the fractional owner becomes the queue described above.
Agathon's Cardinal40 case study describes "A 20-month fractional CTO engagement, completed in 2026." The role was "Fractional CTO owning product requirements, technical architecture, team recruitment, product roadmap, and development cadence from day one." The product is in beta with enterprise users.
A fractional Head of AI, beside the CTO you keep
This is the route the rest of this piece covers: a part-time owner of the AI decisions who works alongside your CTO and leaves the rest of technology with them.
A full-time hire, when AI needs someone every day
This route is covered further down, after the sections on what a fractional Head of AI does and how to judge one.
When a fractional Head of AI makes strategic sense
Agathon describes the Fractional Head of AI role as follows: "I own the AI function only: strategy, model and vendor choices, AI team capability, and the judgment calls in between. The right shape when you already have a CTO or engineering leads and need senior AI ownership alongside them, not above them." The fractional model gives the business specialised leadership without a full-time executive hire. It makes particular sense when:
- Your engineers build AI features that work, and nobody decides which ones matter to the business.
- AI demos run well in a meeting and then stall before customers or staff use them.
- Model and vendor choices are made by whoever is building, with no plan for the next year or two.
- Your CTO is the obvious owner of these decisions and has no time for them.
Same depth, different scope
Agathon's page states the difference between its two fractional shapes in one line: "Same person, same depth, same accountability. The difference is scope." A fractional AI CTO owns all of the technology. A fractional Head of AI owns the AI calls, while the CTO keeps the platform, the wider engineering team, security and delivery.
The test: will your CTO hand over the AI calls?
This shape depends on two answers from the CTO. Do they want the AI decisions taken off them? And will they accept someone else owning those calls inside their organisation? If both answers are yes, the CTO gets time back and the AI decisions get an owner with time to make them. If either answer is no, two people would each believe they decide the same question, and there is reason to expect friction. A CTO who wants to keep the AI decisions and can find the time is better served by the advisory route.
What a fractional Head of AI does
The job is to make the AI decisions and leave the team able to make them without help. In practice that means:
- A plan for how the AI parts fit together and connect to your existing systems.
- Build or buy calls, and a check on whether the AI tools you already pay for earn their cost.
- Written rules on which data the AI may use and how it is tested before customers see it.
- Coaching the CTO and engineering leads on AI questions, so the decisions stay with the team when the role ends.
How to judge a candidate
Look for technical depth on models and architecture, and the ability to explain those calls to a board. An interview on its own can miss both. Better tests:
- Have them design one of your planned AI features with your engineers in the room.
- Have them explain the same plan to your board without jargon.
- Ask what they decided on a past AI project and what they would change now.
- Ask what the business should do when the AI gets something wrong.
If nobody inside can judge the answers, the advice below on who judges full-time candidates applies here too.
Signs the arrangement is working
Look for changes you can check without reading code:
- Each AI decision has a named owner and a written reason.
- New AI features reuse what earlier ones built.
- AI tools the business already pays for are in use or cancelled.
- The CTO spends less of the week on AI decisions.
Hiring a full-time Head of AI
When AI becomes a standing function, with its own team, budget and daily decisions, a part-time owner becomes the constraint, and the business needs someone in the role every day.
Write down the decisions before the job description
Before drafting a job description, write down the decisions the hire will own: which vendors they choose, which budget they control, which work they can stop, and who they report to. The list exposes gaps a job title hides, and it helps the business avoid hiring a researcher when it needs an operator.
Decide who will judge the candidates
If nobody inside can assess depth in AI, a confident candidate and a capable one look the same in interview. Decide who will judge the candidates before the role is advertised. Agathon's advisory page covers this: "I also help with AI hiring, the Head of AI included: defining roles, reviewing candidates, conducting technical interviews and advising on offers."
The shift from fractional to embedded AI leadership
Fractional leadership should make itself unnecessary by building internal capability. Agathon's service page is explicit: "Handing over to a full-time CTO is the planned end of an Agathon Fractional AI CTO engagement." The transition can run in three phases:
- Direct leadership: The fractional leader makes key decisions and directs AI strategy
- Collaborative leadership: The fractional leader works alongside emerging internal leaders
- Advisory oversight: The fractional leader provides periodic guidance to capable internal leaders
Plan and measure this transition explicitly, with clear milestones for handing over responsibilities. For the later question of when to move from fractional to an in-house team, there is a separate piece on that sequence.
A test to run this week
Write down every AI decision the business has to make in the next six months, including vendor and model choices, build-or-buy calls, hires and pilots that might need stopping. Be specific: "choose a document-processing vendor by March" is a decision, while "AI strategy" is not.
Against each line, write who decides it today and how many days a month it needs from its owner. "The CTO, in theory" and "whoever is building it" are both useful answers.
If every line has an owner with authority and time, stretch the CTO and add independent advice. If the lines lack an owner with authority but need only some days, go fractional: a Head of AI if there is a CTO who will hand over the AI calls, and an AI CTO if there is no senior technical leader. If the days add up to a full week every week, hire full time, and use the list as the first draft of the role.
FAQ
What is the difference between an AI consultant and a fractional Head of AI?
An AI consultant recommends and someone in the business decides, while a fractional Head of AI owns the AI decisions part-time and answers for them. Agathon's Fractional Head of AI holds decision-making authority over the AI function, working alongside an existing CTO or engineering leads.
What is the difference between a fractional AI CTO and a fractional Head of AI?
A fractional AI CTO owns all technology decisions, while a fractional Head of AI owns the AI function beside an existing CTO or engineering leads. Agathon offers both shapes, and its service page puts the difference in one line: "Same person, same depth, same accountability. The difference is scope."
How many days a week does a Fractional AI CTO work?
An Agathon Fractional AI CTO typically works one to three days a week, depending on your needs and complexity. I block those days for your company and commit to them at the level of a permanent hire.
How is AI leadership advisory different from hiring a Head of AI?
AI leadership advisory is an external advisory role, while a Head of AI is an employee who runs the AI work day to day and is measured on it. Agathon's leadership advisory gives your leadership team an independent read on the decisions a Head of AI would carry out, with pattern recognition from other organisations and no stake in the answer. I also help with AI hiring, the Head of AI included: defining roles, reviewing candidates, conducting technical interviews and advising on offers.
How does a fractional CTO hand over to a full-time CTO?
Handing over to a full-time CTO is the planned end of an Agathon Fractional AI CTO engagement. I help define the role from what I've learned about your needs, take part in the search, conduct interviews and manage the transition. The engagement has done its job when you no longer need me.






