What does an AI consultant actually do?

What an AI consultant does day to day

Strip the title back and the job is decision support. A business has more AI ideas than time, most of the ideas are mediocre, and the cost of picking the wrong one is a few months and a chunk of budget. The consultant is the person who has seen enough of these to tell the difference quickly.

Mapping the work. Before anything else you have to know where the hours go. That means walking through how a quote gets written, how a job gets scheduled, how an invoice gets chased, and how many people touch each step. Most of the useful findings surface here, and a fair few of them have nothing to do with AI.

Choosing what goes first. A shortlist of two or three candidates, judged on how often the work happens, how much time it eats, how rule-based it is, and what it costs when it goes wrong. Anything that fails those tests goes on the later list, with a note saying why.

Picking tools. Named tools with reasons attached, not categories. Whether Claude or ChatGPT suits your kind of work better, whether you need a team plan or a couple of individual seats, whether n8n or Zapier fits the automation you have in mind, and whether any of it beats software you already pay for.

Sequencing and scoping. What comes first, what has to be true before the next step starts, and what finished looks like in a sentence a non-technical person can check. Without that last part you get projects that run forever because nobody agreed what done meant.

Translating. A good chunk of the value is sitting between the business and whoever builds, so the brief says what the business actually needs and the quotes that come back can be compared against each other.

Getting people to use it. Rollouts fail on habits far more often than on technology. That means time spent on how someone will actually use the thing at 3pm on a Tuesday, not a slide deck about capability.

Here is how that compares to the two other routes people take.

What an AI consultant does not do

Ask whether implementation is included, rather than assuming it comes with the advice. Advisory and delivery are different businesses with different economics, and quietly mixing them is where conflicts of interest live. I keep mine apart on purpose: Consulting covers advice, and if you want something built that is a separate engagement with its own scope, price and contract.

An AI consultant is also not a data scientist, and for most businesses under a few hundred staff that is fine. The work in front of you is almost never a custom model. It is a language model writing first drafts, an automation moving data between two systems nobody has connected, and someone deciding which of those to do first.

They cannot fix a broken process by putting AI on top of it either. If quoting is slow because three people have to approve it and two of them are on site all week, that is a process problem. Automating the emails around it makes a slow process slightly faster and considerably harder to change later.

And they do not sign off on legal, tax or financial questions. A consultant can tell you what a tool does with your data and which questions to put to your lawyer or your accountant. The answer has to come from the person qualified to give it.

The clearest case for skipping a consultant entirely: you already know exactly what you want built, and you have tested the idea enough to be sure it is worth the money. Go straight to an agency. Advice at that point is a tax on a decision you have already made. The expensive version is the reverse, hiring a builder while you are still working out what to build.

When hiring one makes sense

The clearest signal is a decision you keep postponing. There is a quote sitting in your inbox for a build you cannot evaluate, or three tools on trial and no way to choose between them, or a team that has quietly split into the people who use AI daily and the people who opened it once.

The second signal is the size of the waste. If a process eats ten hours a week across your team, an hour of advice that redirects it is cheap at almost any price. If it eats twenty minutes a month, run your own experiment and keep the fee.

The third is speed. Most owners reach the right answer eventually. What you are buying is the months in between, and for a business trying to hold a position those months usually matter more than the fee.

It does not make sense if you have nothing written down and no real sense of where the time goes. Track that for a fortnight first, then book. That gives us a useful starting point for our conversation.

What it costs to hire one

AI Consulting is A$1,195 per month + GST, with a call each fortnight, clarity on what to build in Claude or ChatGPT and practical guidance on how to do it. No lock-in.

Custom development is A$150/hour + GST, with a separate project scope, estimated hours and terms before work starts.

We agree the work, estimated hours and engagement terms before starting.