AI Consulting vs an In-House Team

Considerably more than the salary, which is the number everyone anchors on. Take the advertised base — experienced AI and machine learning engineers in Australian capital cities are commonly advertised somewhere in the $120,000 to $200,000-plus range, though that varies enormously by city, seniority and how loosely the title is used. Then add the superannuation guarantee, which is 12% from 1 July 2025. Then payroll tax if you are over your state’s threshold. Then annual and personal leave, meaning you get roughly 47 working weeks for 52 weeks of cost. Then recruitment, which for specialist technical roles commonly runs 15-20% of first-year salary through an agency. Then equipment, software, cloud and model costs. Then the management overhead — someone senior spending real hours directing them. Then the ramp-up period, because nobody is productive on day one in a business they do not understand. A fully loaded first year is meaningfully above the headline salary, and it is a fixed cost that does not flex if the work turns out to be smaller than you thought.

It is the risk nobody prices, and it is larger than the cost risk. AI is a field where the title is doing a lot of work and the market is full of people whose experience is genuine but adjacent — a data scientist who has built models but never shipped one into a business process, or an engineer who has integrated an API but never diagnosed a workflow. Both are competent people. Neither may be what you need. The problem is that if you do not already have AI capability, you cannot reliably assess AI capability at interview, which is a genuine circularity rather than a failure of diligence. A bad specialist hire in a small business costs a year — a year of salary, a year of the opportunity, and a year of your organisation concluding that AI does not work here, which is the expensive part. The way out of the circularity is to know precisely what the work is before you hire against it, and that knowledge is much cheaper to acquire than a year.

More often than a consultancy would like to admit, and in four fairly identifiable situations. First: when the capability is strategically central rather than supporting — if AI is going into your product rather than your back office, that belongs inside the company permanently and no external arrangement substitutes. Second: when the pipeline is continuous. A team needs a queue of work to justify itself; if you have three years of it, hire. If you have one project, you are hiring a person to finish and then be underutilised. Third: when deep institutional knowledge matters more than technical depth — knowing why your process has that strange step is worth more than knowing three model architectures, and that knowledge does not transfer. Fourth: at real scale, where the coordination cost of external parties exceeds the overhead of employing people. The honest summary: consulting is better at starting, in-house is better at continuing, and most businesses need both in that order.

It works, and it is what most of our clients end up with, because it puts each party where they are actually strong. The external partner does the diagnosis, ships the first workflows, and establishes the patterns — the part that benefits from having done it in twenty other businesses and knowing which twelve things fail. The internal person or team owns it, learns from being alongside the build rather than reading a handover document, and takes over the ongoing work. The reason this sequencing matters is that it inverts the common failure: hiring a team before anyone has decided what they are for, and watching them spend eighteen months building a platform nobody asked for. Diagnose, ship one thing, learn what the work genuinely requires, then hire against that knowledge with a job description written from evidence rather than from a market survey. The hire is better because the specification is real.

Often yes, and it is the most underrated option on this page. Your existing staff have the thing no consultant and no new hire has: they know how the business actually works, including the parts that are undocumented and the parts that are irrational for historical reasons. That knowledge is worth more than technical depth for most first projects, and the technical half is more learnable than the industry lets on — the tools have improved dramatically and the barrier now is much lower than it was even two years ago. The realistic caveats are two. It needs genuine protected time, not "on top of your existing role", which is how these initiatives quietly die. And it needs someone with the temperament for it: curious, persistent, comfortable when something does not work. If you have that person and you can free up their time, back them, and use an external party for the specific things they get stuck on rather than for the whole job.

Deliberately, and with an exit. The free initial consultation and the AI Opportunity Audit at around $3,000 exist to answer "what is the work?" — which is exactly the question you need answered before you can sensibly hire against it. The audit is a written map you keep: workflows, systems, data, constraints, and a ranked shortlist with realistic effort. Plenty of clients take it and do the work themselves or with an internal hire, and that is a legitimate outcome rather than a failed sale. Where we build, we build alongside your people rather than around them, and the handover is a real one — documentation, access, and someone who can maintain it. We are a small independent firm, which means we are the wrong choice for some engagements, and there is a size of business where the honest advice is to hire rather than engage us. We would rather say that in the free conversation than three months in.