---
title: "Who helps a small business put AI to work?"
author: Garrett Winder
updated: 2026-11-19
url: https://garrettwinder.com/put-ai-to-work/
---

# Who helps a small business put AI to work?

An engineer who sits with your team, picks one job someone does every week, and gets AI doing it with the software and data you already have. Small companies rarely have that person on staff. I do this work at your office in Parker County or west Fort Worth, for $2,000 to $6,000 a month, month to month.

## Why do most AI projects fail?

Most of them begin with a technology and go looking for a use. RAND’s 2024 report puts it this way: “By some estimates, more than 80 percent of AI projects fail,” which it says is [twice the rate of IT projects that don’t involve AI](https://www.rand.org/pubs/research_reports/RRA2680-1.html). S&P Global surveyed more than 1,000 companies in 2025 and found the share abandoning most of their AI projects before putting them to use [had risen from 17% to 42%](https://www.spglobal.com/market-intelligence/en/news-insights/research/2025/10/generative-ai-shows-rapid-growth-but-yields-mixed-results). Neither is a study of small companies.

In small companies I see the same few causes. The data the AI needs is spread across several programs and spreadsheets, so it works from an incomplete picture. No one is responsible for the result, so errors go unnoticed and good output goes unused. And the people who do the job weren’t asked, so the AI copies the official procedure when the team worked out a better one long ago.

It also helps to settle the process first. If two people handle the same task differently, choose one way before handing any of it to AI.

## What should AI do first?

A job that comes up often, is tedious, and is easy to check. Good examples are sending incoming email to the right person, writing the first draft of a routine document, copying figures from one program into a weekly report, and answering the questions your team asks each other again and again.

I look for three things. The job comes up at least once a week. Someone can tell at a glance whether the result is correct. A mistake costs a few minutes to fix and nothing more. A job that passes all three is a good place to begin.

Leave out anything where an error could lose a customer or create legal trouble. Those can come later, and a person should approve each one.

I’d also have the AI write drafts before it makes decisions. A person reads each draft and sends it. When the drafts regularly go out unedited, you can let it send the routine ones by itself, and by then the team has seen enough of its work to trust it.

## Do we need new software for this?

Usually you don’t. The AI has to be able to get at the records in the software you already use, and most current products give you a way in. Where one doesn’t, I build that first.

Most of the remaining work is deciding what the AI may do. I set what it can read, what it can change, who approves what it does, and where its activity is recorded. That takes longer than picking a model and has more to do with whether the project works.

What a small company is usually short of is a person to do that fitting. The Census Bureau reported in 2026 that [37% of firms with 250 or more employees use AI, against fewer than 20% of firms with four or fewer](https://www.census.gov/library/stories/2026/05/ai-use-businesses.html). SaaStr has written that most small businesses [expect AI to work out of the box](https://www.saastr.com/forward-deployed-engineer-what-it-takes-to-make-ai-work-in-b2b-but-do-they-work-for-smbs) and have nobody on staff to fit it to how they operate.

I’d skip the AI add-on for each product you own. You’d end up with several assistants, and none of them can see the others’ data.

## Should we build our own software with AI?

Yes, as long as the process comes first: start with strategy, work out who does what and when, then build on tried and true design patterns. AI has made the building cheap. I use it to write much of the code myself, which is why one engineer can now do what used to take a small team, and why software built around your company is within reach.

What AI lets you skip is everything before the building. It builds whatever you describe, and most owners describe everything they know. The tool that comes out holds the whole company on every screen, and the people it was meant for keep working the old way. I’ve written about that in [why the team isn’t using the software you built with AI](/built-with-ai-team-wont-use-it/).

If you’ve already built something, keep it. What you put into it is the hard part, and it stays. The work that’s left is watching how a job moves from one person to the next and rebuilding the screens around that.

## What does this look like in a mineral company?

In most mineral companies the jobs I’d look at first are reading courthouse documents, drafting offer letters from your own valuations, checking incoming revenue statements against what each well should pay, and answering where a deal stands without asking around the office. People usually do all of these by hand, using data the company already has.

The large industry vendors are adding AI to their title and revenue products, and some of it works well. Those tools can’t know how your company values a tract, which owners you’ve already spoken to, or what you offered last time and why. That information is in your files and in your staff’s heads.

So I’d begin with a small project: collect your deal history and owner contacts in one place where an AI can read them. Then I’d give it a single job, with a landman or buyer reviewing the output. Later jobs take less work because the data has already been gathered.

## Next step

I have room for three more companies. $2,000 to $6,000 a month, month to month. It starts with an hour at your office, at no charge. Show me how the work gets done, and I’ll tell you what I’d do next. [Email me](mailto:gw@garrettwinder.com) to pick a day.

---

By [Garrett Winder](https://garrettwinder.com/index.md), who builds custom software for small businesses in Parker County and west Fort Worth, Texas.
