Insights · September 29, 2026
By Mason Hughes
The biggest return I've gotten from AI didn't come from drafting or cutting costs. It came from using AI to surface the information behind decisions. Over the past 44 days, I've run about 7 billion tokens through ChatGPT and Claude while building and running our own business AI system, and most of them went into reading, not writing.
A token is a small piece of a word, about three quarters of one, so that works out to roughly 5 billion words read and written by AI for one small business. Those tokens went into our files, our numbers, our notes, and our processes. Seeing that ratio changed how I think about where AI actually pays off.
You've probably seen the headline that most small businesses use AI now. A 2026 Goldman Sachs survey of small business owners in its 10,000 Small Businesses program found 76% using it. Only 14% said it was fully built into their core operations. And when owners named the main benefit, 84% said efficiency and productivity. Cost cutting wasn't the story.
That matches what I've seen from the inside. Here are the three things I learned.
AI is good at drafting. Emails, outlines, first passes at a proposal: it saves real time there, and you should use it for that.
But the biggest jump I've seen in what AI gives back doesn't come from drafts. It comes from using AI to surface information before you make a call. What did this client actually spend with us over the last two years? Which jobs ran over, and why? What did we promise in that contract? The answers already exist inside your business. They're just scattered across inboxes, spreadsheets, and people's heads.
When AI can pull those answers together in minutes, you decide faster, and you decide with real information instead of gut feel. You aren't replacing anyone here. You're giving the people you already have a much longer reach.
There's a catch. If you're going to make decisions from what AI surfaces, you have to be able to trust it. So we build quality checks around the information itself: every answer that feeds a decision has to show where it came from, it gets checked against that source, and a person signs off before anything changes. That habit came straight out of using AI for decisions every day, and it's now part of how our system works.
AI can create a surprising amount of waste, and most of it starts the same way: buying a subscription without a defined use and without a baseline understanding of how to use it.
Defective drafts. A draft that's mostly right can take longer to fix than writing it yourself would have. Now multiply that across a team.
Rabbit holes. It's easy to spend a weekend building something clever with AI that you never end up using the way you pictured.
Chasing the new thing. Every week brings a new model, a new feature, a new agent. When you're first starting, innovation for its own sake is the most expensive path. Start from a real problem you already have and a possible solution to it. The innovation can come later, once the basics are paying off.
If you're at the subscription-and-prompts stage, the first move isn't a better prompt. It's better information going in.
Connect a database of clean data. Pick one area of the business, like your client list, job history, or price book, and get it into one clean, organized place the AI can reach.
Use formats AI reads well. Plain text, simple documents with clear headings, and spreadsheets with one header row and consistent column names all work well. Scanned PDFs, photos of tables, and spreadsheets full of merged cells and color coding are hard for AI to read accurately.
Understand how it finds information. A basic AI model works from two places: what it learned in training, and what you hand it in the conversation. When you connect your files, it searches them for the pieces that look relevant and pulls those in before it answers. Clean, clearly labeled information gets found. Messy information gets missed or misread.
Get those three right and you'll start seeing higher quality, more tailored outputs, because the AI is finally working from your business instead of the internet's average.
This week, pick one decision you make regularly and write down the information you wish you had in front of you when you make it. Gathering that into one clean place is your first real AI project.
If your team is already using AI without clear habits or boundaries, that is the work behind our AI Training, Coaching & Policy Review.
Source: Goldman Sachs 10,000 Small Businesses Voices survey, conducted by Babson College and David Binder Research, January 27 to February 4, 2026.
20 Questions · 5 Minutes
45 Minutes · Free