Before you build AI, find out if you should.
A fixed-price AI feasibility assessment from senior Australian engineers. We research what's genuinely possible with your data, prove the hard part with a lightweight proof of concept, and hand you a costed plan — or an honest no. Either way, the work is yours to keep.
either way.
Where are you starting from?
Most people arrive at an AI assessment from one of two places. Both are fine starting points; they just need different work.
"We think AI could do this — is that actually true?"
You have an idea, and no way to know whether it's plausible, whether your data can support it, or what it would cost. We research what's genuinely possible for your case, audit the data you'd need, and where the question can't be settled on paper we build a lightweight proof of concept against your real data to prove or kill the hard part.
"We know roughly what we want — how do we actually do it?"
The use case is clear enough; the path isn't. We map where AI fits into how your team already works, what has to change around it, which parts should be automated and which should stay human, and what the sequence of delivery looks like. You get an implementation plan, not a maturity model.
What does an assessment cost?
Fixed-price and quoted up front. Most assessments land around $10,000; a single well-defined question can be $5,000, and planning AI across a whole operation runs to $20,000. That's the only commitment we'll ask you to make today.
Feasibility review
From $5k
Desk research and a data audit on a single use case. Answers "is this plausible, and do we have what it needs?"
Feasibility + proof of concept
~$10k
The review, plus a lightweight proof of concept run against your real data to test the riskiest assumption.
AI implementation plan
$20k
Multiple workflows or a whole operation: where AI fits, what changes around it, and the delivery sequence.
If the assessment says build it, you'll get a costed estimate for the build with your plan. Builds themselves typically start around $50k — but that's a decision for after you know it works, not before.
An assessment is a deliverable, not a workshop.
You're paying for evidence, not enthusiasm. Every assessment produces the same five things, and one of them might be a recommendation not to proceed.
We start from the decision, not the technology
What are you actually trying to change — a cost, a bottleneck, a judgement call someone makes 400 times a day? AI is a means here. Getting the problem stated precisely is what stops an assessment turning into a survey of everything that's currently fashionable.
We audit the data honestly
This is where most AI ideas live or die, and it's the step consultancies skip. Do you have enough data? Is it labelled, consistent, and representative of the cases that matter? If the answer is no, you need to know that now — not six months into a build.
We research what's genuinely possible
What the current state of the art can and can't do for your specific problem, what comparable work exists, and which approach fits — retrieval, a trained model, computer vision, or nothing at all. Written for your team, not padded with vendor material.
We prove the hard part
Where paper can't settle it, we build a lightweight proof of concept against your real data and test the riskiest assumption directly. Often that isn't a working application at all — it's the smallest piece of code that answers the question. Not built to impress; built to find out.
You get a costed plan, or an honest no
What to build, in what order, what it costs, what it should return, and where the risks sit. If the evidence says don't build it, you get that instead — with the reasoning, so you can defend the decision internally.
THE MOST VALUABLE ANSWER MIGHT BE NO
An enormous amount of AI money in the last few years has gone into things that were never going to work — usually because nobody checked the data early enough. If that's your case, the cheapest possible way to find out is a few weeks and a fixed fee, not a six-month build. We'd rather tell you now.
Fixed price. No obligation. Yours to keep.
We won't promise you'll like the answer — the answer is the thing you're paying for, and if we could guarantee it in advance it wouldn't be worth buying. What we will commit to is everything around it: a price agreed before we start, output you keep regardless, and no expectation that you build with us afterwards.
The fee is fixed up front
Scope and price are agreed before the assessment starts and don't move mid-engagement. No hourly meter, no discovery-by-invoice.
You keep everything
The research, the data audit, the proof of concept and the plan are yours — whether you proceed, shelve it, or take it to another firm.
No obligation to build
The assessment is a standalone engagement. Plenty of clients take the output to their board first. Some decide not to proceed, which is a perfectly good result.
Thirty years of shipping software, not slideware.
There is a lot of AI advice on the market right now, and most of it has never had to survive contact with production.
Advice from the people who'd build it
Most AI consulting ends at a document nobody can execute, because the people who wrote it never have to. Ours is constrained by buildability — if we recommend it, we can build it, and we're the ones who'd have to.
We'll tell you if the answer is no
We're not selling you an AI programme. A well-argued "don't build this" saves you a six-figure mistake, and it's a perfectly acceptable outcome of an assessment we've been paid for.
We've actually shipped this
Computer vision in production across six councils and 600,000+ analysed collections, and models reading field instruments for a global testing business. Not a practice we opened when AI got popular.
AI we've actually put into service.
Two AI systems running in the real world, doing work that used to need a person looking at every case.
Your data, your model, your call.
Anyone serious about AI asks these three things first. Here's our answer before you have to ask.
Your data stays yours
We'll tell you exactly where your data goes, what touches it, and what it's used for. Nothing gets fed into training a public model, and we'll put the boundaries in writing before we start.
You own the model and the plan
The research, the proof of concept, any model weights, the plan — yours, whether or not you build with us. Take it to your board, your internal team, or another firm.
Accuracy gets measured, not claimed
We'll tell you how good it actually is on your data, how we measured that, and what happens on the cases it gets wrong. An AI system without an error plan isn't finished.
Find out whether it works.
Tell us what you're considering. We'll come back with an honest read on whether it's worth assessing at all, what an assessment would cost for something your size, and whether we're the right people to run it.
- Fixed-price assessment, typically around $10,000
- You keep the research, the proof of concept and the plan either way
- No obligation to build — a reasoned "no" is a valid result
- Senior Australian engineers with production AI in service
Prefer to talk? Call us on 03 9955 7000