The AI Users Who Hired More People

Say you run a small cabinet shop. Six people, a full workshop, and a folder of estimates you meant to finish over the weekend.

On Monday, a customer calls to ask whether you still want the job. You do. You just haven't priced it yet. Another customer wants a change to a drawing. Someone on the floor needs a decision before cutting the next piece. By lunch, the folder hasn't moved.

Ask where you need help and you might say another person in the office. Ask what you would do with more confirmed orders and you might say another cabinetmaker.

Both answers could be right. The order matters.

That is what makes a new piece of small-business research interesting. Some owners using AI appear to be reaching the second problem: finding more people to do the work.

A different story showed up on payroll

In research published Thursday, September 10, payroll company Gusto followed 2,262 businesses: 1,593 reported using AI, and 669 knew about it but hadn't adopted it. Over the following year, AI adopters had roughly 7% higher headcounts relative to comparable non-adopters. The difference was about 10% for firms with fewer than ten employees.

Among businesses that added jobs, hires included technicians, cooks, teachers and care staff. Those are people doing the work customers buy.

The study has a clear limit. These were Gusto customers who answered a survey, and businesses chose whether to use AI. The findings show an association; they don't establish that AI caused the hiring or predict what will happen at your company.

Still, this is a useful reason to reconsider an assumption. A tool that reduces someone's paperwork can leave the business needing more of their practical skill.

The work can get stuck before it reaches the workshop

Back at the cabinet shop, imagine that preparing an estimate gets easier. An assistant assembles a first draft from your approved pricing notes and job details. You still decide whether the measurements, materials and labor allowance make sense. But you start with something to review instead of an empty page.

If that works, more estimates may reach customers while they are still interested. Some will say yes. Eventually, the question could shift from when you can get a quote out to when your team can build the cabinets.

At that point, another skilled pair of hands might be more useful than another person copying measurements between documents.

The original problem hasn't vanished into a vague claim about productivity. You can follow what changed: a quote left the office, a customer approved it, and a real job reached the floor. That sequence gives you something concrete to measure.

It also explains why hours saved and jobs removed are different calculations. Taking several administrative tasks off someone's plate does not automatically remove the need for that person. It changes what they have time to do.

More estimates are useful when someone wants them

There is a catch in the cabinet-shop example. Customers were already waiting.

If the phone has gone quiet and the order book is thin, producing estimates faster may not be the most useful thing to fix. You could end up with a beautifully organized office and the same empty workshop. A tool can help you respond to demand; it can't make every offer worth buying.

So before choosing an AI project, look at the work that is waiting and ask why it is waiting.

A customer who wants to buy but hasn't received a quote points to one problem. A quote nobody accepts points to another. An approved order waiting for materials points to a third. They might all feel like being busy, but they call for different decisions.

For the shop with a real estimate backlog, try improving one part of that process. Watch whether quotes go out sooner, whether customers accept them, and whether the workshop can deliver what was promised. Include the time spent correcting drafts. A faster first version is only useful if the whole job becomes easier.

Then decide whether there is enough repeat business to support another person. That decision belongs to the order book and the cost of doing the work, not a national AI headline.

Tell the team what the saved time is for

An owner might see room to grow. The person preparing estimates might hear that their job is about to disappear.

You can avoid some of that uncertainty by being specific at the start. In this example, the aim could be to clear the estimate backlog and give the office more time to speak with customers before orders reach production. That is a much clearer assignment than telling everyone to become more efficient.

It also gives the team a useful test. If the new process creates extra checking, missing details or promises the workshop cannot keep, they should say so. Those observations help you judge whether it is doing the job you intended.

And if the payoff turns out to be getting home on time with the same team, that counts too. Growth is one possible use of recovered time. An owner does not need to hire someone to prove a tool was worthwhile.

Start with the folder

For the owner in our example, the next sensible move is small: find out how much work is waiting for an estimate, and why preparing it takes so long. If a tool can remove part of that delay without adding mistakes, try it on a few jobs you understand well.

Then watch where the work goes. Does it become an order? Can the team handle it? Is the pressure moving toward the work you would happily hire someone to do?

Those answers will tell you more about your next hire than another prediction about AI replacing everyone.

If you want to work through which part of your business is holding up the next job, we can look at it together.