The first hire your business should make now that AI exists

SME owners ask me about AI tools all the time. Which one, how much, how quickly. Almost nobody asks the question that actually matters more: how does AI change the next person we hire?

It should change it a lot. A 12-person business hiring its 13th person in 2026 is making a meaningfully different decision from the same business making the same hire in 2022. The roles that needed a human three years ago aren't necessarily the ones that need one now. And the roles that didn't exist three years ago are quietly becoming the most valuable ones on the team.

Here's what that actually looks like when you're sitting down to write a job description in a world where AI has already changed what "full-time equivalent" means.

How does AI change who I should hire next?

The short answer is that AI has changed the economics of three things: production, coordination, and judgement. Those shifts should change your hiring priorities.

Production work (writing a first draft, building a deck from a brief, pulling numbers from a database, drafting a response to an email) has become dramatically cheaper. The 2023 Harvard-BCG study of 758 consultants is the best-known data point here: for tasks inside the frontier of what current AI can do, participants using GPT-4 were 25 per cent faster and produced work rated 40 per cent higher quality than the control group. That's not a small business running a marketing agency, I know. But the pattern holds lower down the scale, and the pattern is what matters.

Coordination work (scheduling, summarising meetings, routing messages, keeping track of what's been done) is in a similar position. Meeting transcription alone gives most managers an hour or two back in a typical week. Multiply that across a team and you're looking at a meaningful amount of labour freed up.

Judgement work has barely moved. Deciding which client to say no to, reading the room in a difficult conversation, knowing when a brief is wrong before the work starts, understanding what a half-formed complaint actually means. These are the things AI is still bad at, and will stay bad at for a while.

What that means for your next hire is simple. The value of a human who spends most of their day on production has dropped. The value of a human who spends most of their day on judgement has gone up. If the role you're about to advertise is mostly the first kind, you're probably hiring the wrong role.

What's the "AI-native hire" everyone's talking about?

You'll hear people describe the ideal 2026 SME hire as "AI-native." It's a slippery term. What it usually means, when you cut past the jargon, is someone who:

Uses AI tools without being asked, and uses them well. Not someone who's done a half-day course. Someone who'll use ChatGPT or Claude or Copilot to think through a problem the same way they'd use a spreadsheet. It's a reflex, not a skill.

Can do the work of someone more senior with AI assistance. A genuinely AI-native mid-level marketer can produce work that would have needed a senior marketer two years ago. Not perfectly, but credibly. The gap between junior and senior output is narrower than it used to be, which matters a lot for SME hiring.

Is comfortable with the parts of their job being automated away. This is underrated. The best AI-native hires aren't threatened by automation of their own tasks. They expect it, sometimes even engineer it themselves.

The problem with hiring for "AI-native" as a label is that the market is flooded with people who will say they're AI-native in an interview and then revert to copy-pasting into ChatGPT once a week when they start. The way around this is not to ask if someone uses AI. It's to ask them to show you how they use it.

More on that in the interview section below.

Which role should I actually hire next?

The honest answer depends on where your biggest constraint currently sits. But for most SMEs in the 10 to 50 person bracket, the three roles worth looking at hardest are these.

The operations generalist who's good with systems. Not a traditional ops hire. Someone who understands how your business runs, can see where the friction is, and has the temperament to systematise it. AI tools make this person dramatically more valuable than they used to be, because the barrier to automating a process has collapsed. Someone who can spot that you spend four hours a week manually reconciling two systems, and then build something in a weekend that handles most of it, is now worth what a specialist engineer would have cost three years ago.

The senior individual contributor, not the junior team. Traditional SME hiring logic says to hire two juniors instead of one senior, on cost grounds. That logic has weakened. One senior person with AI assistance can now produce something closer to what a senior-plus-junior pair used to produce. If you're choosing between one senior at £65,000 and two juniors at £32,000, the senior has become a better bet in more situations than it used to be. Not all. More.

The person who owns AI inside the business. Not a CTO. Not a prompt engineer. Someone whose job it is to keep track of what the team is using, what's working, what's being wasted, and what to try next. In a 30-person business this is a 20 per cent allocation on top of someone's existing role, not a dedicated hire. In a 50-person business it probably is a dedicated hire. Most businesses in this bracket don't have this person and the absence shows up as scattered tool sprawl, forgotten subscriptions, and team members who never get past the basics.

What's missing from this list is as important as what's on it. Pure production roles (write this, design this, schedule this, reply to this) are the ones AI has hit hardest. If the job description is mostly "produces the output," it's probably the wrong hire.

How do I interview for AI ability without being gimmicky?

Most interviews for "AI ability" are useless. They either ask the candidate to describe their "AI experience" (which everyone now knows how to perform) or they run a test that's so artificial it tells you nothing about real-world use.

Three questions that actually work.

The first: "Walk me through a piece of work you did recently where AI helped you get to a better answer than you would have reached on your own. I want to see the actual conversation or the actual document, not a summary." This is the most powerful single question in an AI-era interview. People who genuinely use AI have dozens of examples and can pull one up without hesitation. People who don't will stall, or produce something that's obviously been assembled for the occasion.

The second: "What's an AI tool you tried that didn't work for you, and why?" Anyone using AI seriously has a graveyard of tools they gave up on. The specificity of the answer tells you a lot. Vague dislike ("it wasn't very good") is a bad sign. Specific frustration ("I found it hallucinated names in citations and I couldn't trust it for research") is a good one.

The third: "Here's a real task we need doing. Walk me through how you'd approach it, and where you'd use AI in the process." You're looking for someone who doesn't default to either extreme. The candidate who says "I'd get AI to do it all" is naive. The candidate who says "I wouldn't use AI for this" is probably out of date. The candidate who says "I'd use it here, here, and here, but I'd write the X bit myself because that's the part that matters" is the one you want.

Don't bother with standardised AI skills tests. They don't measure what you care about.

What about existing team members?

A lot of SME owners read articles like this and conclude that their existing team needs to be quietly replaced. That's almost always wrong, and usually expensive.

The better question is what existing team members could do if they had proper AI support and proper training. Rough experience: most existing SME employees become significantly more capable with 10 to 15 hours of genuine AI training and a clear expectation that they'll use it. A smaller group takes longer but still gets there. A smaller group again either can't or won't adapt, and that last group is usually smaller than most owners expect.

What this means practically: before you hire anyone new, work out whether the role you think you need actually needs a new person, or whether it needs an existing person with new tools and new training. The answer isn't always the same, and the decision is worth taking seriously. A badly hired role is more expensive than a good training programme by a factor of three or four.

So what should I actually do about the next hire?

Three things, in sequence.

Before you write the job description, spend a day thinking about which parts of the role AI will do most of in 18 months. Write the description for the role that will still exist then, not the one that exists now. If you can't think of a version of the role that will still exist in 18 months, that's a significant signal.

When you interview, ask the three questions above. Weight them more heavily than the standard competency questions you'd have asked in 2022. AI ability isn't the only thing that matters, but it's now more predictive of on-the-job performance in most SME roles than three or four of the things it's probably displacing in your interview panel.

And before you sign the offer, run a sense check with yourself. If this person produced only 70 per cent of what I'd expect and spent the other 30 per cent of their time improving how we work, would that still be a good hire? For the roles worth filling in 2026, the answer should be yes. If it isn't, you're probably still hiring for the old world.

Businesses that think carefully about this tend to end up with smaller, more senior teams producing more than their bigger predecessors did. Businesses that don't tend to keep hiring the same roles they hired in 2020, keep paying roughly the same for them, and then sit in their offices wondering why the productivity gains everyone else is talking about haven't shown up on their P&L.