Nearly every business owner weighing up AI arrives at the same question sooner or later: does getting good at this mean needing fewer people.

A survey published this week by Small Business Expo puts a number on it for the first time in a way that is genuinely useful. Roughly one in four small business owners have already replaced, or say they would replace, an employee with AI. That is a real number, and worth sitting with rather than dismissing.

The more telling number sits right next to it. 45.2% of small business owners, the largest single group in the survey, say they would only replace an employee with AI if it became necessary (source: Small Business Expo, 2026 Small Business Owner Survey, published 14 September 2026). Not never. Not as soon as possible. Only if necessary.

That is a healthier starting point than either extreme doing the rounds in public conversation about AI and jobs. It is neither "AI is coming for every role" nor "AI changes nothing about how a business is staffed." Most small business owners are treating headcount as a last resort when AI enters the picture, not a first move.

The same survey found something else worth noting: nearly seven in ten small business owners are using, or considering, debt to fund AI investment, and return on investment, not novelty, has become the deciding factor in what gets bought. That is not the behaviour of people chasing a trend. It is the behaviour of people making a considered financial decision and expecting it to earn its keep, which fits the same pattern as the headcount finding: careful, not reckless, and not naive either.

What "only if necessary" looks like in practice

The honest answer to the headcount question is rarely a single yes or no. It is closer to: look at where the friction actually is before deciding anything about people.

In practice that means working through the same ground a proper AI Accelerator Diagnostic covers before a staffing decision gets made, not after. Where is AI genuinely taking friction out of the work, and where does it just move the friction somewhere else. What does a role actually still need once its repetitive parts are handled by a tool, because the honest answer is rarely "nothing." And how will anyone know, three months on, whether a change worked, rather than guessing from how busy things feel.

Skip that groundwork and a business ends up in one of two unhelpful places: cutting a role too early because a tool looked capable in a demo, or sitting on a decision for a year because nobody wants to be the one who gets it wrong. Neither is necessary. The survey suggests most owners already sense this. "Only if necessary" is exactly the discipline a proper look across the six dimensions, vision, opportunities, people and culture, tools and data, governance and value, is built to support.

A fuller picture, not a single headline

It is worth reading this alongside data on small businesses that have actually leaned into AI. Payroll data published earlier this month by Gusto showed that small businesses actively adopting AI grew headcount around seven percent faster than those that had not, with the effect stronger still among businesses under ten employees. Put the two findings together and a clearer, calmer picture emerges. The small businesses doing best with AI right now are not the ones cutting staff fastest. They are the ones working out, deliberately, where a person is still the right answer and where a tool is.

That is a far less dramatic story than "AI replaces jobs," and a much more useful one for anyone actually running a business. The headcount question deserves a considered answer, built on a proper look at where the friction is and what the business needs once it is gone, not a reaction to whichever headline landed first this week.

Go wisely.