The biggest AI opportunity may be buried inside people who don't consider themselves technical

The expertise was already there

Take an accountant.

They understand their clients, tax, financial statements, compliance, billing, the exceptions that cause problems and the repetitive work that somehow still happens every month.

They might not know Python. They might not know what an API is. They might have no interest whatsoever in learning how a database works.

But they know that every month somebody spends hours reviewing work in progress, comparing time recorded against what has been billed, looking for anomalies and working out which clients need attention.

That knowledge is enormously valuable.

Previously, turning that knowledge into software required another person to understand the accounting workflow well enough to translate it into requirements, then someone else to build it.

Now the accountant can start the conversation themselves.

They can explain how the process works today. They can give AI examples of the information involved. They can ask it to identify repetitive steps, decisions, exceptions and areas where judgement is required.

Before anyone writes a line of code, they can start shaping what a better process might look like.

That's a very different starting point.

The same thing applies to a lawyer

A lawyer might spend enormous amounts of time reviewing documents, comparing clauses, researching legislation, preparing advice and producing variations of documents they've created many times before.

The opportunity isn't necessarily to tell AI:

"Be a lawyer and do all of this for me."

That's probably the least interesting interpretation of what's happening.

The lawyer already has the expertise.

The opportunity is to use AI to extend it.

Compare these clauses and show me the differences. Find the parts of this document that relate to a particular obligation. Help me structure the first pass of this research. Check whether I've considered the obvious counterarguments. Turn these notes into a draft I can review. Give me the source for this proposition so I can verify it.

The lawyer remains the lawyer.

But the amount of ground one lawyer can cover changes considerably.

And importantly, they know what a bad answer looks like in a way that somebody outside the profession doesn't.

That matters.

Or take someone who doesn't sit behind a desk at all

Imagine an electrician running a small business.

They probably don't wake up wanting an AI strategy.

They might finish a job, drive to the next one, answer a customer call, photograph some work, order materials, send a quote that night and then remember three days later that an invoice still hasn't gone out.

There's plenty AI could potentially help with there.

Turn rough job notes into a professional customer summary. Draft a quote from a description of the work. Help prepare a scope. Categorise incoming enquiries. Produce follow-up messages. Turn a voice note made in the ute into something useful. Help create safety documentation or explain a technical document in plain English.

A plumber could use it in much the same way. So could a builder, a landscaper or someone running a cleaning business.

They don't need to think of themselves as technical.

They need to understand their work.

That's the part AI doesn't bring to the table.

Domain knowledge just became more powerful

I think that's one of the more interesting shifts.

We've spent years treating technical capability and domain expertise as two fairly separate things.

The accountant understood accounting.

The developer understood software.

The electrician understood electrical work.

The consultant or business analyst often sat somewhere in the middle, translating between the person with the problem and the people or systems capable of solving it.

Those roles aren't disappearing. There are plenty of problems where specialist technical capability is absolutely required.

But the distance between the domain expert and a working solution has become much shorter.

That's been my own experience.

I've worked around technology, software platforms, solution design and business processes for a long time, but I wasn't a traditional software developer.

AI has allowed me to take what I already understood and push much further into areas I previously would have needed other people to execute.

PAR-18 started with playing golf with a mate and keeping score in a spreadsheet. That eventually became an actual iOS application.

Performance Signal started because I was taking a ridiculous number of screenshots from Garmin and giving them to ChatGPT after a run. That eventually became a way for ChatGPT to work directly with the underlying running data.

Neither idea came from sitting around trying to invent an AI product.

They came from knowing the problem.

That's the bit I think organisations should pay attention to.

Your best AI idea might not come from the AI team

When businesses think about AI adoption, there's a temptation to start centrally.

Pick the approved tools. Establish the governance. Run some training. Find a few use cases. Build a roadmap.

Some of that is necessary, particularly around data, security and responsible use.

But I wouldn't assume the people running that program know where all the valuable opportunities are.

The person doing the same reconciliation every Friday probably knows something they don't.

So does the person who spends two hours turning meeting notes into a client report. Or the person who knows exactly why a particular approval process always gets stuck. Or the lawyer who has reviewed the same type of agreement 500 times. Or the accountant who knows which five pieces of information they're forever chasing from clients.

Those people have spent years developing an incredibly detailed understanding of the work.

What they often haven't had is an accessible way to do much about it.

Now they do.

Which makes upskilling much more interesting

This is why I think AI education inside organisations needs to go further than showing people where the chat box is and giving them ten useful prompts.

People need to understand what these tools are capable of.

They need to understand where AI is strong, where it isn't, what information it can work with, how to challenge its answers, when something needs checking and when a deterministic system or conventional software would be more appropriate.

But then you need to give them room to explore.

Ask people what annoys them about their job.

What do they repeat constantly?

What takes far longer than it should?

What information do they have to pull together manually?

What do customers keep asking them?

What would they investigate more deeply if it didn't take half a day?

What have they always thought surely there must be a better way to do this?

That's where I'd start.

Because once someone understands what AI can actually do, they begin looking at their own work differently.

"I'm not technical" matters a lot less than it used to

There are obviously limits to this.

Being able to ask AI to build software doesn't suddenly make someone a software engineer. Using AI for legal research doesn't remove professional obligations. An accountant still needs to understand the advice they're giving. An electrician shouldn't trust a generated answer over the standards and professional knowledge governing their work.

That human expertise becomes more important as the tool becomes more capable, not less.

But I think we're going to discover an enormous number of people who have been sitting on valuable ideas without ever thinking of them as technology ideas.

They understand the customer.

They understand the workflow.

They understand the exceptions.

They understand what good looks like.

They've just never been able to do much with that knowledge themselves.

AI changes that.

The biggest opportunity may not be teaching technical people how to use AI.

It might be showing millions of people who don't consider themselves technical that, for the first time, they don't necessarily need to be.