Your staff already use AI at work. Train them before you buy software.

The first money a small company spends on AI should go to training its people and writing one page of rules. It should not go to a new platform. We say this as a firm that builds software for a living. Your staff are already using AI, with or without a decision from you, and that fact should set the order of what you do.

More people use it than you think

Gallup reported on July 20, 2026 that 52% of American workers now use AI in their role. 30% use it a few times a week or more, and 15% use it every day. In the same survey, 47% said their organization has put AI tools in place. One in five said they do not know if their organization has done so.

Those numbers do not prove that people are hiding anything. They do show that more workers use AI than work at a company that has set it up. A fifth of workers cannot say what their employer’s position is. The usual name for this is “shadow AI”: staff using AI tools that the company did not choose and may not know about.

For a company with no technology team, think about what that means on a normal day. Someone in the office pastes a customer’s details into a free chatbot to draft a reply. Someone else uses a different tool to summarize a contract. Nobody decided any of this, so nobody checks the results or knows where the data went.

Buying a platform does not fix it

The common advice is to buy a company-wide AI product so that everyone uses the same approved tool. A reasonable person can hold that view. We think it gets the order wrong.

The owners themselves say that software is not what they lack. In a Goldman Sachs survey from March 2026, 76% of small businesses said they use AI, but only 14% said it is fully part of their core operations. 73% said they would benefit from more training and help with putting it to use.

The San Francisco Fed read through what small business owners wrote about AI in a national credit survey, and posted the findings on September 1, 2026. The answers were collected in 2024, so they are not new. Owners who wanted to use AI named the cost of tools, the time and capacity it takes to train staff, and not knowing how to start. A new platform adds to the first two and does nothing about the third.

There is also a reason to think skill is where the return is. In the Gallup survey, 45% of workers who use AI for one or two purposes said it had improved their productivity. Among those who use it for seven or more purposes, 90% said so. Gallup did not test what causes what. But it fits what we see: the gain comes from people who know several ways to use the tool, and a new license does not teach anyone that.

What it looked like at one company

A startup we worked with had nobody to do its marketing. People there had already tried AI assistants, and it did not help much. Every new chat started from zero. Someone had to explain again what the company does, how it sounds and what the logo looks like. Posts still got written only when someone had a spare hour.

We did build software for them in the end. But look at how we knew what to build. The plain tools had already shown where they failed. The system keeps the company’s voice, colors and logo in one place, so nobody explains them twice. It also follows one firm rule: the AI writes a draft, and only a signed-in person can approve it or publish it.

That is the order we recommend. People learn the tools, the company writes down its rules, and then, if one task still eats hours every week, you build or buy something for that one task.

One page of rules

The rules do not need a lawyer or a consultant to start. They need to answer three questions that your staff are already answering alone.

Which tools may we use for work? Name them. If the answer is “the free version of whatever you like”, write that down and see how it looks on paper.

What never goes into them? We would start with customer records, payroll, and anything covered by a contract or a privacy law. Your list may be different, and your staff need to know what is on it.

Who checks the result before it leaves the building? When an AI draft goes to a customer unread, the company made that mistake. The tool did not.

Then train people on their real work. Leave the general introduction out. Use the quote they wrote on Tuesday and the report that is due on Friday. Small businesses are moving this way already. The U.S. Chamber of Commerce reported on October 2, 2026 that 66% of small businesses now use AI, and that 95% of those are working to build their employees’ skills, including 47% that train on the job.

After all that, the page of rules and the trained staff will tell you which software is worth paying for. We expect it to be less than a vendor would sell you.

If you want your team trained on its own daily work, book a 30 minute call and we will plan a session with you.