← All field notes

Should you hire an AI lead or an AI consultant? A guide for small businesses

Most businesses of 10 to 50 people don't need a full-time AI lead. They need the right person for a few months, a clear idea of what success looks like, and a plan for who carries it after that person leaves.

If you run a business of 10 to 50 people, you probably don't need a full-time AI lead. You need the right person, often an AI consultant on contract, for three to twelve months, and a plan for who carries the work once they've gone.

I'll be upfront: I sell this kind of work, so weigh what follows with that in mind. These are the questions I'd want a client to answer before they hire anyone, including me.

Write down what success looks like first

Start with the job, not the person. What should be different in 12 months because this person was here?

There are two very different answers. One is finishing systems you've already started and never got working. The other is rebuilding how the business runs with AI at the centre. They need different people, different budgets and different timeframes.

US and UK recruiters keep seeing the same pattern: companies hire an AI leader before they know what the role is for, and the search restarts within a year. If you can't write the outcome in two or three sentences, you're not ready to hire yet.

Match it to the business you are now, too. The Tech Council of Australia found only 25% of Australian SMEs have systems and data ready for AI, and 88% of those using it struggle to connect it to what they already run. If your data lives in spreadsheets and inboxes, that's where the work starts.

Size and complexity decide the shape of the role

The bigger the company, the more likely you need someone full time. A team of 10 to 20 might need someone for two or three months. Smaller than that, shorter again.

Headcount isn't the whole story. A 15-person firm with three practice areas, a compliance regime and five systems that don't talk to each other is more complex than a 40-person business doing one thing well. Size the role to the complexity, not the staff list.

Talk to your people before you hire anyone

Has your company actually talked about what AI means for the people in it? Whose jobs change when it's in place? If it's everyone, or a specific group, what's the plan for them?

Implementing AI isn't the only work here. Some of your team will worry about their jobs, and some will push back quietly because they can see what the changes might mean for them. An AI lead walking into that without the conversation having happened will spend months fighting it.

Jobs and Skills Australia's view is that AI is more likely to change work than replace it, but people need bringing along. Trust is the other half. Around 65% of Australian SMEs that hold back from AI say distrust of its decisions is the reason. Decide early where a person checks the output before it reaches a customer.

Business background often beats a tech-only CV

This one I learned the hard way. On a recent engagement I realised that, despite my startup background, I didn't know the standard processes every services business runs on. My startup years were about growth and getting a product to market, which is what programs like YC's Startup School teach. A services business works very differently, and I had to learn it on the job.

So ask what your candidate's background is. Software engineering or systems admin only? Or have they worked in, or with, a business like yours?

Someone who already understands services processes can sit with each department head for an hour and get what they need. Someone who doesn't is learning two things at once: how your business works, and where AI could help it. Do you have the patience and the support around them for that learning curve?

The people who know both aren't cheap, to put it politely. I know people who charge $1,000 a day. Decide where you want that money to go.

Engineering problem or science problem?

Look at each project on the list and ask which kind it is.

An engineering problem has been solved before. Business automation existed long before the current wave of AI, and tools like Make, n8n and Cflow already do a lot of it. You can estimate the time and the cost with some confidence.

A science problem is newer. Much of what people want from AI right now has no standard way of doing it yet. You need clear goals, a time limit, and an honest answer to "how long are we willing to try before we decide it doesn't have legs?"

This changes whether you build at all. MIT's research on generative AI pilots found tools bought from specialist vendors or delivered with partners succeeded about 67% of the time, while internal builds succeeded about a third as often. For a small business, buy or partner first, and save the building for the parts that are truly yours.

Security and the rules are part of the job

Whoever you bring in needs enough engineering or IT knowledge to see how a change affects your existing setup and your security. That's becoming a major headache, not a side issue.

Most SMEs haven't set this up yet. The Tech Council found only 26% have rules on which AI tools staff may use, and 21% have none at all. Privacy and cyber security are the top two worries. The federal government's Guidance for AI Adoption includes an AI policy template, an AI register and a screening tool built for small businesses, which is a good place to start. Make writing those down part of the AI lead's job.

Test with proof, not a CV

Ask candidates to show you something they built and explain what changed because of it. Have them pitch the business case to your least technical manager, then ask that manager whether they understood it. Call a former colleague and ask what was different after they left.

Anyone can list tools on a CV. What you're hiring is judgement about your business.

Plan the handover from day one

Do you already have an IT person who looks after the day-to-day? They've probably been too busy keeping things running to learn AI, and that's not their failing.

Writing software is much less of a bottleneck than it was, because AI does a lot of it now. What's scarce is someone who understands your systems and your people. So bring an outside lead in for three, six or twelve months to learn your setup, build what's needed and document it well enough for your IT person to pick up. That person has been with you for five, 10 or 15 years. They shouldn't fall behind on AI just because they've been busy holding the business together.

Don't stop at one person, either. The Tech Council found 39% of SME employees had very little AI training and 26% had none. Everyone who'll use the new systems needs some of the handover, not only IT.

The questions to answer before you hire

  • What should be different in 12 months, written in two or three sentences?
  • Are we finishing systems we started, or rebuilding around AI?
  • How complex are we, not only how big?
  • Have we talked honestly with the people whose work will change?
  • Where does a person check the output before it reaches a customer?
  • Does the candidate know businesses like ours, or will they be learning on our time?
  • Which projects are engineering problems, and which are science problems?
  • What can we buy or partner on before we build?
  • Who writes our AI rules, and who keeps them current?
  • Who takes over when the AI lead leaves, and how will we train everyone else?

How I'd do it

For most businesses I talk to, I'd bring someone in for a fixed period with a clear outcome, a handover plan and an internal person named from the first week. A full-time AI lead makes sense once you're big enough that there's always the next project waiting.

If you'd like to talk through which one fits your business, book a free intro call.


Sources: Tech Council of Australia, Accelerating AI Impact for SMEs (October 2026); National AI Centre, AI adoption insights (May 2026); Department of Industry, Science and Resources, Guidance for AI Adoption (October 2025); Jobs and Skills Australia, Our Gen AI Transition (2025); Section, You're probably hiring the wrong Head of AI (September 2026); Christian & Timbers, The 5 biggest mistakes when hiring AI leaders (August 2025); Workplace Journal on La Fosse (May 2026); Fortune on MIT NANDA's generative AI pilot study (August 2025).

If you read this far

Two things you could do next.

Get This Week in Claude in your inbox every Friday morning.

Your email stays with me. Unsubscribe in one click.

Rather have it set up for you than piece it together yourself?

Book an intro call

Free, 30 minutes. No obligation.