Your Team Already Built Your Best AI Tool

A six-person HVAC company outside Paterson. Every morning the service coordinator spends the first forty minutes turning overnight voicemails into a route. Listen, write the address on a pad, check the map, decide who gets what.

One Sunday he got tired of it. He opened a chatbot, described the problem, and by dinner he had a little page where he pastes the transcripts and gets back a grouped route by zip code with the emergencies flagged at the top.

He has used it every morning since. Nobody asked him to build it. The owner found out in September, by accident, when the coordinator was out with the flu and the mornings stopped working.

The useful stuff is already in the building

New numbers from Pave, the app-building product inside Quickbase, put a size on this. They surveyed 5,177 employees and 527 business leaders in September 2026. Among employees, 38% used AI to build an app or working prototype in the past year, and 24% built something they actually use for work. Strip out the people in IT, software and security and the number barely moves: 22% of everyone else built a tool they use.

Worth noting that Pave sells an app builder, so read their framing with that in mind. The underlying pattern still holds up, and it matches what I see in client offices.

Two details in that survey matter more than the headline. The first is what these tools replace. 84% of them target a process that used to live in a spreadsheet, an email thread, a chat, or on paper. These are not science projects. They are the exact friction you complain about on the drive home.

The second is who asked for them. Out of 1,970 people who built something, nine did it because a manager or colleague requested it. Nine. Everyone else built it because the task annoyed them. 64% said the whole point was saving time or cutting manual work.

Nobody is going to hand you a proposal for this. It is already happening in the gaps between tasks.

You probably think you know about all of them

Now the uncomfortable part. In the same survey, 70% of leaders said they felt confident they had a complete picture of the employee-built apps in their department.

In that same group of organizations, 53% found at least one app in the past year that leadership and IT never knew existed.

Both numbers cannot be right. The confidence is the problem, not the apps.

In a company of six or twelve or thirty, there is no IT department to route this through, which means there is no moment where anyone writes it down. The tool just exists, inside one person's browser, doing real work. Only 14% of the people who built these tools said their organization had designated anyone to maintain or support them. Only 32% had access to someone who could look at it for security or risk.

So the thing holding up your Monday mornings has one author, no backup, and no review. That is not a reason to panic. It is a reason to go find out what you have.

The instinct to shut it down is the wrong one

Most owners hear all this and reach for a ban. The data says do not.

When leaders in the Pave survey discovered an employee-built app, 83% called it important or essential to the process it supported. Fewer than 2% made the employee stop using it. And 80% said they want employees to stay involved in building and running this stuff.

That tracks. Your coordinator knows the route problem better than any vendor ever will. He is not guessing at the workflow, he lives in it. That is the part consultants charge you to learn.

The gap is not enthusiasm. It is that almost nobody has set the boundaries. Goldman Sachs and Babson College surveyed 2,473 small business owners in August and found that 74% of the owners who use AI have no formal policy at all for how their employees use it. Just over half, 56%, are doing anything on upskilling, and most of that is handing people a tool and hoping.

You do not need a policy binder. You need two sentences and a list.

What to do with one afternoon

Five things, in order. None of them require a consultant, and you can get through all five before close of business.

Ask the question out loud. At the next team meeting: has anyone built or rigged up something to make their own job easier? Say plainly that nobody is in trouble and nothing is getting taken away. You will be surprised what comes out, and you will only get honest answers if you promise that first.

Write down what exists. One line per tool: what it does, who built it, what information goes into it, and what breaks if it stops working tomorrow. A single page is enough. The page is the whole point, because right now it is in nobody's hands.

Put a second person on anything that matters. If the morning route dies when one person has the flu, that is not a tool, it is a hostage situation. Walk one other person through it this week.

Set the two rules that actually matter. What information never goes into an outside AI tool, naming the specific things: customer payment details, employee records, anything under a client NDA. And who signs off before anything AI-generated goes to a customer. Two rules beats twelve.

Make one of them official. Pick the tool with the most value and the most fragility, and spend real money making it solid: proper access, a place it lives that is not one person's laptop, and someone whose job includes keeping it alive.

Your best AI project is already paid for

The thing most owners want from AI is a system built by someone who understands exactly how the work moves through their business. You already employ those people. Some of them started building it months ago without telling you.

The question is not whether to let them. It is whether you find out before the flu does.

If you want help taking that inventory, or turning one of these homemade tools into something the whole team can lean on, that is the kind of work we do at Nexera Intelligence.