Fractional AI Lead vs Head of AI: Which Do You Need?

Hire a Head of AI when AI is core to your product and the role will still be full-time in three years. Take a Fractional AI Lead when you need senior judgement now, the workload is real but not yet forty hours a week, or you cannot credibly interview for a role you have never held. The distinction is not seniority or cost. It is whether the work is permanent and whether you can currently evaluate the person doing it.

Most mid-sized companies get this wrong in the same direction. They open a Head of AI search because that is the legible move, spend six to nine months not filling it, and make architecture decisions in the meantime by default rather than by choice. This article is about how to tell which situation you are actually in.

What does a Head of AI do that a Fractional AI Lead does not?

A permanent Head of AI builds and owns an organisation. They hire, they set career paths, they hold budget across years, they sit in the leadership meetings where priorities are traded, and they accumulate institutional context nobody external can match. If your AI capability needs to grow to a team of eight and stay there, that is the job, and no fractional arrangement substitutes for it.

A Fractional AI Lead owns decisions rather than an organisation. They set the roadmap, make architecture and vendor calls, define the evaluation standard, hold the regulatory posture, and are accountable for delivery. What they do not do is build a department or carry political weight in a leadership team they only partly belong to. Where the constraint is direction, that is sufficient. Where the constraint is organisational growth, it is not.

How long does hiring a Head of AI actually take?

Longer than the plan assumes, and the gap is where the damage happens. Between writing the specification, searching, interviewing, notice periods, and the months before a new leader is productive in an unfamiliar company, nine to twelve months from decision to effectiveness is a realistic figure for a senior AI hire in a mid-sized European business. Senior machine learning leadership is genuinely scarce, and the candidates worth having are rarely looking.

The cost of that window is not the empty salary line. It is that the first architecture decisions get made anyway, by whoever is available, and a wrong one compounds. A vendor gets selected, a data pipeline gets built a particular way, a pilot gets scoped against no threshold, and the incoming Head of AI inherits commitments they would not have made.

Can you interview well for a role you have never had?

Usually not, and this is the least discussed failure mode. Evaluating a senior AI candidate requires distinguishing someone who has shipped and maintained production systems from someone fluent in the vocabulary. Without that experience on your side of the table, the interview selects for articulacy, and articulacy is abundant in this field.

A fractional arrangement resolves the problem in an unglamorous way: the person doing the work writes the specification for the permanent role, sits on the panel, and can tell the difference. Several of the better outcomes we have seen run exactly that sequence, with the fractional lead deliberately hiring their own replacement.

Which is right when AI is core to the product?

A permanent hire, without much hesitation. If the model is the product, or a substantial part of it, then AI decisions are product decisions, they happen daily, and they need someone whose incentives are wholly aligned with the company over years. Fractional leadership is a poor fit for work that never stops.

The nuance is sequencing. Even product-led companies frequently benefit from fractional coverage during the search itself, precisely so the architecture decisions made during those nine months are deliberate. That is a bridge, explicitly framed and time-boxed, not a substitute.

Which is right when AI supports the business but is not the business?

This is where fractional usually wins, and it describes most manufacturers, healthcare providers, professional services firms and financial institutions. AI matters, sometimes considerably, but it is a capability that improves operations rather than the thing customers buy.

In that shape the workload is genuinely lumpy. Intense during an audit and a first build, much lighter during steady-state operation, intense again when a system needs retraining or a regulation changes. A permanent senior hire is under-occupied through the troughs and difficult to justify at review time, which is how good people end up leaving roles that were never sized correctly. Our post on what the fractional role actually owns covers the mandate in more detail.

What about hiring a strong engineer instead?

A common and expensive substitution. A capable machine learning engineer will build what they are asked to build, competently. What they typically will not do is decide whether it should be built, challenge a use case that will not survive contact with production, negotiate scope with a sponsor, or own the regulatory position.

The result is a well-engineered system solving a problem nobody validated. That failure is more expensive than a slow hire, because it consumes a year and produces a working thing that changes nothing. If you can only fund one senior person and the roadmap is unclear, judgement is the scarcer input.

How do the two compare on risk?

They fail differently, which matters more than which is cheaper.

• A wrong permanent hire is slow and expensive to unwind: months of underperformance, a difficult exit, and a second search from a worse position. In some European jurisdictions the notice and process obligations make this substantially harder than the hiring conversation implies.
• A wrong fractional engagement ends at the end of a contract period. That is a genuine structural advantage while your requirements are still uncertain.
• The fractional-specific risk is under-mandating. A fractional lead without real decision rights is a consultant with a different title, and the arrangement produces advice nobody is accountable for. If the mandate does not include architecture sign-off, that is the failure mode to expect.
• The shared risk is no internal owner. Whichever you choose, if nobody internal owns the system in month nine, it decays. A fractional engagement should end with that person named.

Can you run both at once?

Yes, and it is more common than the framing suggests. A fractional lead alongside an existing CTO or Head of Engineering is a normal arrangement: the CTO owns the platform and the engineering organisation, the fractional lead owns the AI roadmap, model decisions and regulatory posture.

It works when the boundary is written down before the engagement starts rather than negotiated during it. It fails when both roles believe they own architecture, which surfaces at the first genuinely contested decision and is unpleasant for everyone.

How do you decide?

Four questions, answered honestly, resolve most cases.

• Will this be full-time work in three years? If yes, hire. If the honest answer is that you cannot tell, that uncertainty is itself an argument for the reversible option.
• Can you evaluate the candidate? If nobody internally has shipped production AI, your interview is measuring the wrong thing.
• What happens in the next nine months if you do nothing? If the answer is that decisions get made by default, you need coverage during the search regardless of the eventual destination.
• Is AI the product? If yes, hire, and use fractional cover only as an explicit bridge.

If the answers point to fractional and you are not yet sure what should be built, an AI Readiness Audit is the cheaper first step. It produces the roadmap that tells you how much leadership capacity you actually need, which is the input both options require and neither supplies.

The honest summary

Hire a Head of AI when the work is permanent, evaluable, and central to the product. Take a Fractional AI Lead when the work is real but not yet permanent, when you cannot credibly interview for the role, or when the nine-month search would otherwise be nine months of decisions made by default. The comparison that misleads people is day rate against salary. The comparison that matters is the cost of a wrong architecture decision against the cost of senior judgement, and on that basis the arithmetic is rarely close.

The related question of consultant versus lead, which is about accountability rather than permanence, is covered in Fractional AI Lead vs AI Consultant, and the in-house team version of the argument in AI consultant vs in-house AI team.

Frequently asked questions

Is a Fractional AI Lead just a contractor?

No. A contractor executes a defined scope; a fractional lead holds decision rights and is accountable for outcomes across a mandate. The practical test is whether they can say no to a proposed project. If they cannot, the arrangement has been mis-specified regardless of what it is called.

What happens when we do eventually hire permanently?

The fractional lead should write the role specification, sit on the interview panel, and hand over documented architecture decisions and a roadmap. Handing over to a permanent hire is a normal and expected ending, not a failure of the engagement.

Will a fractional lead understand our business well enough?

For roadmap and architecture decisions, generally yes within weeks, because those depend on your data, processes and constraints rather than on long-accumulated relationships. Where fractional genuinely trails a permanent hire is internal political capital, which matters most when the work requires changing how other departments operate.

How much of a Head of AI's job does fractional actually cover?

The decision-making, roughly. Not team building, not headcount planning, not the standing presence in leadership meetings. If you list what you want the Head of AI to do and most items are decisions rather than organisation-building, fractional covers more than you would expect.