A founder asked me last month whether he should hire a Head of AI or bring us in on a fractional basis. He had a salary figure in his head and a vague sense that an employee would be "more committed." Fair questions. The answer depends on numbers he had not yet run, so we ran them together. Here is what that looks like.
The salary comparison, done properly
A Head of AI or AI Lead in an Australian SME is going to cost you somewhere between $180,000 and $240,000 in base salary, depending on city and how much genuine AI experience you insist on. Add superannuation, payroll tax, leave loading, equipment, and the recruitment fee if you use an agency. Call it $230,000 to $310,000 all-in for year one.
That is before the part nobody budgets for. The first 3 months are onboarding. Most of the next 3 are building context. You are 6 months in before the hire is doing the job you hired them for, and you have paid roughly $120,000 to get there.
A fractional AI officer engagement with us runs from $1,500 a month for a lighter-touch arrangement to around $3,500 a month for the full scope. Annualised, that is $18,000 to $42,000. No onboarding period because we do this across several businesses at once and walk in with the patterns already in hand. No super, no leave, no recruitment fee. Cancel month to month.
So the headline is roughly 6 to 1 in favour of fractional. Most people stop there. They should not.
The harder comparison
Price only matters if you are buying the same thing. You are not.
An in-house AI hire in a 20 to 200 person business is one person. That one person has to do governance, which is slow and careful. They have to do strategy, which needs a view across the whole business. They have to do hands-on implementation, which means actually building and shipping. They have to train staff. They have to evaluate vendors without being captured by any of them.
Those are four or five different jobs. The people who are genuinely good at all of them are rare and expensive, and the ones you can afford at SME scale tend to be strong in one area and learning the rest on your payroll. You end up with an expensive governance person who cannot build, or a capable builder who has never written a policy and does not want to.
There is also the single point of failure. When your one AI person leaves, and at that salary band they will be recruited, everything they knew leaves with them. Half-finished roadmaps. Vendor relationships. The reasons behind decisions. You start again.
A fractional arrangement spreads that risk. You get someone who has done the governance piece a dozen times, has built the workflows, has sat in the vendor meetings and knows which claims to ignore. The work is documented because it has to be. The roadmap belongs to you and lives in your systems, not in someone's head.
When in-house is the right call
I will be straight about this because the comparison is not always one-sided.
If AI is your product, hire. If you are building something where the model is the business, you need that capability on the payroll and you need more than one person.
If you are over 300 people with a real technology function, you are probably past the fractional model. At that scale you can afford a team, and a team can cover the range that one person cannot.
If you have already done the governance and strategy work and what you need now is a full-time builder, hire a builder. A fractional AIO is the wrong tool for that job.
For most businesses between 20 and 200 people, though, the arithmetic and the capability argument point the same way.
What the first 90 days actually look like
Since the question underneath "should I hire" is usually "what will I actually get," here is the shape of a typical fractional engagement.
Month one is mostly listening. We map the workflows that eat the most time. We find out where AI is already in use, officially or otherwise, and what that is exposing. We put a short, usable AI policy in front of leadership and get it signed off. We pick two or three quick wins that can be live inside weeks.
Month two is building and shipping those wins, measuring them, and starting the roadmap conversation properly now that there is evidence on the table. Staff training starts here too, on the tools they are actually going to use, not generic AI literacy.
Month three is about proving the return and deciding what is next. Leadership gets a report with real numbers in it. If the numbers are not there, that is a conversation we have openly, and the month-to-month arrangement means you can walk.
By the end of that quarter, the business has a policy, a roadmap, two or three working improvements, a trained team, and a clear view of whether to go deeper. An in-house hire is still in onboarding.
The thing the founder said afterwards
After we went through all this, the founder I mentioned said something I think about a lot. He said the salary comparison had made him feel like he was being cheap by going fractional. The capability comparison made him realise he had been about to spend $280,000 to buy one person's opinion.
Run both comparisons. Then decide.