The Best Person to Use AI Is Someone Who Already Knows the Job

Give the same AI tool the same job, once from a first-year graduate and once from someone with twenty years in the field, and you’ll get two very different results.

Take a builder’s estimator pricing a bathroom renovation from a set of plans. The graduate asks the AI for a cost breakdown and gets a tidy, plausible one, with line items for tiling, fixtures, plumbing and labour. The veteran looks at the same plans, notices the house was built in the 1960s, and asks different questions. Is there asbestos in the wall sheeting? Will the waterproofing need bringing up to the current standard? Can the floor framing take the weight of the stone bath the client has fallen for? The AI answers those questions well too, once somebody thinks to ask them.

The two estimates can end up thousands of dollars apart, and only one of them survives contact with the job. The tool was identical in both cases; the person driving it wasn’t.

This article is about that difference, which shows up well beyond software. In every industry we work with, the people getting the most out of AI already know the job well enough to direct it and check it. The businesses that will do best with AI are the ones that build their own expertise into how they use it.

Why does an expert get more from the same tool?

Mostly because the value of an AI answer depends on the question, and an expert asks better questions. They know which details matter, what usually goes wrong, and which of the client’s assumptions need challenging. A novice asks the question as it was handed to them.

The expert can also tell when the answer is wrong. This matters more than it sounds, because AI tools write with the same fluency whether they’re right or not. A confident, well-formatted answer with a subtle error in it looks exactly like a correct one to someone who doesn’t know the subject. An experienced bookkeeper spots a misapplied GST rule on sight, where a graduate might copy it straight into the client’s return.

And the expert knows what the answer is for. A lawyer drafting a letter knows how this particular client reacts to a firm tone and what happened the last time the other side was pushed. None of that is in the prompt, though it often decides whether the letter works.

We’ve written about the software version of this in what happens when your AI says one thing and your developer says another. The pattern is the same in accounting, law, engineering, marketing, trades and medicine. In each of them, AI is at its best as an assistant to someone who knows the work, and at its riskiest when it’s used in place of them.

What is a business’s “secret sauce”?

Every good business knows things that aren’t written down anywhere. There’s a way it prices a job with an awkward site, and a short list of suppliers who look cheap and turn out not to be. Somebody knows what a particular kind of customer means when they say “urgent”. And the senior person runs a mental checklist before anything goes out the door, built from twenty years of things that went wrong.

That knowledge is the reason clients choose one firm over another. It’s also the part a general AI tool doesn’t have.

The large AI models learnt from what’s publicly available, which means they’re very good at the typical answer, the one an average practitioner in your field would give. Ask them without adding anything of your own and you get the industry’s average answer, faster. Your competitors, using the same tools, get the same one.

Build your own knowledge into how the tool works and it starts producing your answer instead: your pricing logic, your checklists, your standard of work, at a speed nobody could manage by hand.

What does building expertise into AI look like?

It’s less exotic than it sounds, and most of the work isn’t technical.

Write down what your best people know. Expect this to be the hardest step. It’s also where most of the value sits. Sit with your senior people and capture the rules they apply without thinking: how they decide, what they check, what makes them stop and ask a question. Businesses that do this often find it pays for itself before any AI is involved, because it’s also how new staff get trained properly.

Give the tool your material. That means proposals that won, standard clauses, pricing rules, past jobs with notes on what went wrong, your answers to the questions customers ask most. Depending on the need, that might go into a well-configured AI workspace with your documents attached, a searchable knowledge base the AI draws on, or a purpose-built tool connected to your own systems. For some processes it becomes an AI agent that carries out the work within rules you set.

Put your experts at the checkpoints. The AI drafts and an expert reviews, at least until the drafts are reliably good. The expert’s corrections are the valuable part. Each one is a piece of knowledge that wasn’t captured yet, and it should go back into the instructions so the same correction isn’t needed twice.

Judge it by your own standard. “Does it look professional?” is too low a bar, because AI output nearly always looks professional. Take real past cases where you know what good looked like and see how close the tool gets.

Doesn’t AI make expertise less valuable?

It would be dishonest to claim AI doesn’t make some expert work cheaper. The routine drafting, summarising and first-pass research that used to fill a junior’s week can now be done in minutes. That changes how firms staff their work and how they charge for it, and it raises a real question the professions haven’t answered yet: if juniors no longer do the routine work, how do they build the judgement that makes a senior?

But the part of expertise clients pay the most for becomes more valuable. When a business can produce ten times as many drafts, someone has to decide which one is right, and be accountable for it. What clients are paying for is confidence that the advice is correct for their situation, with somebody who knows the field standing behind it.

That points to a strong business model, simple enough to state in a sentence. Sell the judgement and the outcome, and use AI to deliver them faster and more consistently than before. A business that relies on AI without expertise has nothing to offer that its clients couldn’t get from the same tool themselves. A business that combines deep knowledge of its field with AI built around that knowledge can do better work, in less time, than either could alone, in a way that’s very hard for a competitor to copy.

What this means for you

Before choosing an AI tool or starting an AI project, find the thing your best people know that your newer people don’t. Write it down. That document is worth more than any subscription, because every useful AI system in your business will be built on it.

Then look for the place where that knowledge is applied most often, at the most volume, with the most waiting around. That’s usually where AI built around your expertise will make the biggest difference to your clients.

Simon Paul is a Business Solutions and Technology Specialist at Code Brewery who’s spent 25+ years turning business ideas into software that actually earns its keep. He has watched plenty of tools arrive promising to replace experience, and he’s still waiting to meet one that could. Reach out to Simon to talk through how your business’s know-how could shape the way it uses AI.