Judge the method, not the demonstration. Every agency you speak to can show you something impressive. Far fewer can explain, in specific terms, how they decide what to build and what happens when the first thing they try doesn’t work. That’s what separates the firms that deliver from the ones that present.
Fair warning: I run one of these, so discount this accordingly. I’ve written the version I’d want if I were the one buying, including the parts that don’t flatter firms like mine.
Why is this a commercial decision?
Because when an engagement goes wrong, your business carries it.
A stalled engagement wastes more than the fee. It ties up your people, leaves half-integrated processes nobody trusts and makes the next attempt harder, because the business now has a story about AI not working here. That lasts longer than the invoice.
Writing in Harvard Business Review in November 2025, Jin Li, Feng Zhu and Pascal Hua put the pattern plainly: firms struggle to capture real value from AI “not because the technology fails” but because “their people, processes, and politics do.” Keep that in mind when the people you’re evaluating are selling technology.
So most of the selection criteria that matter aren’t technical. Can they scope tightly? Will they tell you when something’s a bad idea? Does the work survive them leaving? Those questions predict outcomes better than any model comparison, and they’re harder to fake in a first meeting than a demonstration is.
The tools are the cheap part, and everyone already has them. You’re paying an agency for judgement about what to do first, and for someone accountable for the result. If the output’s wrong in month three, someone has to care.
Most businesses have plenty of plausible places to use AI. A few will pay back quickly. Others will be technically feasible and commercially pointless, and some will do harm, usually by automating a judgement that should stay with a person. Telling them apart in advance is most of the value, and it’s hard to do from inside the business, where every process feels either obviously broken or obviously fine.
What separates capability from a good pitch?
Four things, none of which show up in a slide deck.
How they choose. Ask an agency to walk you through how they decide which of your workflows to touch first. A strong answer is specific and a bit dull: they’ll want to know what you already measure, where work gets redone and which jobs are frequent and structured. A weak answer talks about possibilities and use cases in general terms. This one question sorts the field faster than anything else.
Someone at a small advisory firm told me how they won a recent piece of work. Their pitch said plainly that they didn’t yet know what the result would look like, then set out the questions they’d ask and how they’d work. Every other bidder pitched an answer. The client chose them, reasoning that nobody can have that answer yet, so the confident pitches were showing a confidence that doesn’t exist. An agency that arrives with the answer before it has asked you anything is telling you how it chooses.
Whether the work outlives them. Ask what your team can run and change after the engagement ends. Some arrangements are built so you can’t maintain anything without the agency. That’s a deliberate business model. Choose it knowingly, or you’ll discover it later.
Ownership. Every automated process needs a named person on your side who’s accountable for the output being right. If an agency hasn’t raised this by the second conversation, they haven’t implemented much.
What they won’t do. A partner who can’t name anything they’d decline is either inexperienced or selling. The most useful question in the whole process is: where do you think AI is the wrong answer for us?
What are the trade-offs between the different kinds of firm?
Three broad models, each good at something different.
| Type | What they’re strong at | Where they struggle |
|---|---|---|
| Large enterprise consultancies | Complex multi-year programmes, organisations with internal AI teams and formal governance, board-level assurance | Cost and pace. Their delivery model assumes a scale most Australian businesses don’t have, and you rarely get the people from the pitch |
| Technical AI labs and dev shops | Custom model work, novel technical problems, deep engineering | Commercial framing and adoption. Strong builds that never get used, because the workflow and training side was somebody else’s job |
| Focused specialist agencies | Applied work inside existing processes, speed, defined scope, adoption | Depth on hard technical problems. If you need a custom model trained, this is the wrong door |
My read: most firms of 5 to 100 people need the third, occasionally the second and almost never the first. The services a specialist offers should read like a list of your own workflows. But if you’re in a regulated sector with a board that needs formal assurance, the first exists for a reason, and a specialist telling you otherwise is selling.
When would I be better off not hiring an agency?
Sometimes the answer is none of them. Some businesses run on a platform built for their industry that already does most of what they need. I’ve had that conversation with a plumber and with a lawyer, and both were substantially right. If that sounds like you, find out what the software you already pay for can do before you pay anyone to add to it.
Three other situations point the same way.
You already have the person. If someone internal has the mandate, the time and the process instinct, back them. They know your business better than any agency will in twelve weeks. Buy them training and time.
Your first project is obvious and self-contained. If there’s one clear job, the data is clean and one team owns it end to end, just do it. Get help if it doesn’t work. A lot of first automations don’t need outside help, and I’ll tell you so when yours doesn’t.
It isn’t an AI problem. Sometimes what looks like an AI opportunity is an undefined process, an unresolved argument about ownership or a system nobody has configured properly. AI on top of that makes the underlying problem faster and harder to see. Fix the process first. It’s cheaper, and afterwards you may not need the AI at all.
Another plumbing business I spoke to had already done exactly that. They’d brought in a coach, sorted their processes and got the admin off the owner’s desk. I could see places AI would help, but not enough to be confident it would beat what they’d just built. I said so, and we left it there.
What should I ask an agency?
Take these to every conversation. The answers are more revealing than any case study.
- How do you decide what to build first? Listen for whether they ask what you measure today.
- Where is AI the wrong answer for us? Silence here is disqualifying.
- What happens if the first project doesn’t work? Everyone has one. The good ones tell you without being pushed.
- Do you resell or earn margin on any platform you’d recommend? Margin can be fine when it’s disclosed.
- Who owns this internally when you leave, and what can they change?
- What does the scope end look like? A defined end with a fixed fee focuses the mind. Open-ended arrangements remove the pressure to produce anything.
- What have you declined recently, and why?
Expect fair questions back: what you already measure, who’d own this internally and what you’ve tried before that didn’t work. If nobody asks that last one, they aren’t listening.
Why do pilots so often fail to scale?
Because a pilot that works and a process that works are different achievements, and the gap between them is organisational.
A pilot is run by motivated people, on a case they chose, with someone watching. Daily operations have none of those conditions. What closes the gap is standard inputs, a named owner, a set review rhythm, training that shows people their own job and a point where you decide to keep it or stop.
Ask any prospective partner how they handle that transition. If the answer is “we hand it over”, you’ll be doing the hard part yourself. I’ve written more on what moves a pilot into daily use, and you can see how we structure that transition in Audit, Build, Train, Partner.
How much should governance factor into the decision?
More than most buyers think. Less than most vendors sell.
For a business your size, governance means writing down a handful of answers: what customer information can go near these tools, who’s accountable for the output, what happens when something’s wrong and who’s allowed to change the setup. That’s a page or two of plain language.
Two reference points, if you want them. The NIST AI Risk Management Framework and the OECD AI Principles both cover transparency, accountability and risk management, and a competent partner can talk about them without reaching for a brochure. Your page should be shorter and more specific than either.
What should I avoid?
- Choosing on brand. A recognisable name tells you how good their marketing is. Ask who’ll do your work.
- Buying the demonstration. It was built to succeed. Ask what it took to build, and what it would take on your data.
- Trusting an early forecast. Anyone quoting you a percentage improvement before looking at your business is guessing.
- Accepting an open-ended engagement because it feels flexible. It removes the deadline that makes results happen.
- Skipping the reference call. Ask to speak to a client where something went wrong. The refusal is the answer.
- Assuming a proposal is a plan. A document that hasn’t been tested against one of your workflows is a hypothesis.
- Treating price as the variable. The expensive mistake is a year spent on the wrong project.
How should I run the selection?
Keep it short. You don’t need a procurement process.
- Write down the problem, in your own words, before you speak to anyone. “Quotes take three days and we lose work” is workable. “We should be doing something with AI” will get you sold to.
- Speak to three, of different types, so you can hear the difference in how they answer.
- Ask each one the same questions and compare the answers.
- Take a reference call, and ask about something that didn’t go to plan.
- Scope the first piece small enough to judge within weeks, with a fixed fee and a defined end. If a proposed first step can’t show you anything for a quarter, ask why it’s that large.
- Agree what “it worked” means before starting, in a number you already track. The useful measure is usually time or consistency: a quote that took three days goes out the same day.
If a first conversation doesn’t leave you with a clearer view of your own problem, that tells you something, whatever they’re selling. The people worth hiring tend to be useful before you’ve paid them anything.
Ours is a discovery call. Tell us how your business runs, and afterwards I’ll come back with the first step I’d take and what it would cost. The FAQ covers the usual questions, and you can read what clients say in their own words.
FAQ
How do I choose an AI consulting agency in Australia?
Judge how they work. Ask how they decide what to build first, where they think AI is the wrong answer for you, who owns the result internally and what the scope end looks like. Speak to three firms of different types and compare their answers to the same questions.
What questions should I ask an AI consultant before hiring them?
The most useful is where they think AI is the wrong answer for your business. Then: what happens if the first project fails, whether they earn margin on any platform they’d recommend, what your team can run and change after they leave and what they’ve declined recently.
Should I hire a big consultancy or a specialist agency?
It depends on your scale and how much formal assurance you need. Large consultancies suit organisations with internal AI teams and formal governance requirements. Most firms of 5 to 100 people get better value from a specialist working inside their existing processes, because the work is faster, cheaper and more likely to be adopted.
Do I need an AI agency, or can I do this myself?
Plenty of businesses do the first automation themselves, and should. You need outside help when the obvious project is done and the next decisions get harder, or when nobody internally has the time and authority to own it.
How much should an AI consulting engagement cost?
The structure matters more than the number. Look for a defined scope with a fixed fee and a fixed window, and be wary of an open-ended retainer. If the first step can’t be judged within weeks, it’s scoped too large.
What are the warning signs of a bad AI consultant?
They can’t name anything they’d decline. They lead with technology and never ask what you measure. They quote a percentage improvement before they’ve looked at your business. They won’t disclose platform commercial relationships. They resist a defined scope end. And they can’t give you a reference where something went wrong.
