Insights · AI Delivery
What AI consulting costs in Australia.
Nobody publishes a price. Here is what actually drives the number, the four commercial models you will be offered, and the questions that expose an open-ended engagement before you sign one.
Search for AI consulting in Australia and you will find a lot of capability and almost no numbers. That is frustrating if you are the person who has to take a budget to a board, and it is the single most common question we get before a first meeting.
This article does not contain a price, and you should be suspicious of any that does. Not because the number is a secret, but because the honest version of it depends almost entirely on the condition of your data, and nobody can see that from the outside. What we can do - and what is genuinely more useful - is tell you what moves the number, how the four common commercial models shift risk between you and the supplier, and what a quote has to itemise before it is safe to sign.
The four commercial models you will be offered.
Almost every proposal you receive will be one of these, or a blend. The difference that matters is not the headline rate. It is who carries the risk when the work turns out to be harder than the proposal assumed.
- Time and materials (day rates). You pay for effort. The supplier carries no delivery risk, so if the data is worse than expected, the cost of that discovery is yours. Appropriate for genuinely undefined discovery work. Dangerous as a default, because there is no commercial force pushing the engagement towards an end.
- Fixed scope. A defined deliverable for a defined price. The supplier carries the risk of it taking longer, which is why the scope, the assumptions and the variation mechanism are written down carefully. The thing to inspect is not the price, it is what counts as out of scope.
- Outcome or milestone-based. Payment tied to something measurable being true. Attractive in principle, and workable where the outcome is genuinely measurable and mostly within the supplier's control. Where it goes wrong is when the "outcome" is a metric you both influence, and the argument moves to attribution.
- Embedded capacity. Senior specialists added to your team for a period, working to your priorities. Honest, flexible, and the right answer when you know what to build and lack the hands. It is effort-based, so it needs the same discipline as day rates: a defined review point, not an open tap. That is the shape of our forward deployed engineers model.
A reasonable rule: use effort-based pricing where the problem is genuinely unknown, and fixed scope everywhere the deliverable can be named. If a supplier only offers day rates for a piece of work you have described precisely, that is information about the supplier.
What actually drives the number.
In our experience the cost of an AI engagement in Australia is set mostly by five things, and four of them sit on the client's side of the table.
- The state of the data. This is the big one, and it is why estimates move so much after the first week. Data that is already joined, documented and reasonably clean can support a use case quickly. Data spread across systems with no shared identifier means the first real deliverable is a data foundation, not a model.
- How many source systems have to be joined. Cost rises with the number of integrations far faster than it rises with the sophistication of the analytics.
- How long access takes. A month of waiting for credentials and security review is a month of elapsed engagement. This is the cost driver most often left out of proposals entirely, and the one most within your control to fix before day one.
- The regulatory surface. Work touching APRA-regulated operations, health data, or public-sector obligations carries an evidence burden that is real work, not overhead. If you are in that category, our notes on CPS 230 and AI and on what the NSW AI Assessment Framework asks of your consultant set out what the supplier is actually being asked to produce.
- Whether capability has to transfer. Delivering a working thing is one price. Delivering it in a way your team can run and extend without the supplier is a different, higher, and usually better-value price.
Notice what is not on that list: the modelling. The algorithmic part of most Australian AI projects is the cheapest and shortest phase. Almost all of the cost, and nearly all of the risk, sits in the data underneath it.
The eight lines a quote should itemise.
You do not need to be technical to assess a proposal. You need to check that these eight things are written down. If any of them is missing, the gap is where the variation will come from later.
- Named people and their allocation. Not "a senior data engineer" - who, and what percentage of their week. Ask what else they are on.
- Deliverables with acceptance criteria. What you will receive, and how you will both know it is finished.
- Assumptions, listed explicitly. Every fixed price rests on assumptions. A proposal that does not list them has hidden them.
- Dependencies on you. Access, data, decisions, people's time. With dates.
- What triggers a variation, and how it is priced. Agreed before you start, not negotiated when you are committed.
- Third-party costs. Licences, cloud, tooling: passed through at cost, or marked up, and if marked up, by how much.
- IP and code ownership. Who owns what is built, and whether you can take it elsewhere.
- Handover. What documentation and knowledge transfer you get at the end, in writing, as a deliverable rather than a promise.
Three questions that expose an open-ended engagement.
These are the ones that produce the most informative silences.
- "What happens to the price if the data turns out to be worse than you have assumed?" A good answer describes a specific mechanism. A bad answer reassures you.
- "Which named person will be doing this work, and what else are they committed to?" The gap between the people in the pitch and the people on the job is the oldest problem in consulting.
- "What is the smallest first engagement that would still tell us something useful?" A supplier who can answer this is thinking about your risk. A supplier who cannot wants a big first contract.
Why the cheapest quote is usually a scoping problem.
When one proposal comes in materially below the others, the usual explanation is not efficiency. It is that it has scoped less, assumed better data, or left the integration work out. That price is real, but it is the price of a smaller job, and the difference reappears as variations once you are committed and switching is expensive.
The useful comparison is not price against price. It is scope against scope: take the cheapest proposal and the most expensive, and list what the expensive one includes that the cheap one does not. Sometimes the answer is nothing and you have found a bargain. More often the answer is the integration work, the evidence work, or the handover.
How RUBIX prices this.
We work to fixed-scope engagements rather than open-ended day rates, so the cost and the deliverables are known before we start. These are the durations of the four ways clients usually begin - durations, not prices, because the price depends on the drivers above:
- AI readiness assessment - two to four weeks. A prioritised plan before you commit budget. Detail on how we run one.
- Data foundation build - eight to twelve weeks. Core systems joined into one trusted, AI-ready layer.
- AI governance framework - four to eight weeks. Making AI use safe, accountable and defensible. See AI governance consulting.
- Embedded delivery - ongoing, with defined review points.
If you do not know which of those you need, the free AI readiness snapshot takes about ninety seconds and will tell you where the gaps are before anyone quotes you anything. Broader detail on scope and approach sits on our AI consulting services page.
General information only, not commercial, legal or financial advice. Durations are indicative and are not a quote. Current as at August 2026.
