Who Runs Your AI?

A common story goes like this.

Someone at the company starts using ChatGPT. They draft an email, summarize a bid, rewrite a safety memo. It works, and they tell the owner. The owner is impressed and says the whole company should be using it.

Three months later, almost nobody is. A couple of people still use it now and then. Nothing about how the company runs has changed.

The tool didn’t fail. Nobody was in charge of it.

The gap isn’t the software

When a company buys accounting software, someone owns it. There’s a person who sets it up, keeps it clean, trains new hires, and fixes it when it breaks. If that person leaves, the company notices.

AI usually has no such person. It’s treated like a subscription anyone can open, and what’s everybody’s job ends up being nobody’s.

So you get the pattern above: scattered experiments, no shared way of working, no way to tell whether any of it is saving time, and nobody asking whether it’s safe to feed a client’s contract into a chatbot.

What owning it actually means

I think of this person as the AI operator. The title doesn’t matter. The work does:

• Pick the problem. Choose one slow, annoying, repeatable piece of work, such as chasing quotes, writing daily reports, or sorting incoming email, and start there instead of trying to automate everything.
• Set the rules. What can the AI do on its own? What does a person approve? What information never goes in?
• Check the output. AI makes mistakes. Somebody has to look, especially early.
• Measure. Did it save time or not? If not, change it or stop.
• Keep it running. Tools change, staff change, and processes drift.

That’s mostly a coordinator’s job. It needs someone who knows how the company really works, which usually means someone already inside it.

Who could do it

Often it’s not the person you’d guess. A sharp office manager, an estimator, or a project coordinator who already knows where the time goes can be better than a hired technologist, because they know which problems are worth solving.

What matters:

• They understand the work.
• They’re organized and careful.
• They have some time set aside for it. This is the part companies skip. If it’s something squeezed in after the real job, it won’t happen.
• The owner backs them.

One person, or two?

Here is the catch. Knowing the work and building the setup are different skills.

Some setups are easy enough that your key person can do them on their own. Others are too hard for someone who doesn’t program. They involve connecting several tools, or logic that is hard to describe without writing code. In those cases the company needs a second person, someone developer-minded, to take what the key player knows and put it into code.

That second person can be hired or subcontracted. They don’t need to know construction. They need the key player to explain the work clearly, and the key player needs to be able to tell whether the result is right. Neither one can do it alone.

How a company staffs this depends mostly on how complex its setups are, and on its size. I expect a lot of smaller companies to keep someone on retainer rather than hire, because the building work comes in bursts. A larger company may reasonably have a two-person team in house.

None of this changes the first point. The knowledge side still has to come from someone inside the company. A developer can’t decide which problem is worth solving, or whether the output is good enough to send to a client.

What happens if the owner tries to run it alone

This is common, and it rarely lasts. The owner has the most context and the least time. The AI work gets done in bursts, when things are slow, and drops the moment a job gets busy. That’s not a character flaw. It’s what the schedule of running a construction company looks like.

Where to start

Before you buy anything or hire anyone, answer these:

1. Who in the company would be the person? Name them.
2. What is one task they’d go after first?
3. How would you know in 30 days whether it helped?

If you can’t answer the first one, that’s the real starting point, and no tool will fix it. Once you can, the next question is whether they can build it themselves or need a second pair of hands.

A caveat

I’m describing a pattern, not a law. Some small companies do fine with a couple of people using AI informally for personal tasks, and if that’s enough for you, that’s fine. The pattern shows up when a company wants AI to change how the business runs. At that point it needs an owner, and often a builder too.

Want to talk it through?

If you’d like to talk through who could own this at your company, Book a free one-on-one marketing session. I’ll tell you honestly where I’d start.

Benjamin