Almost every small business owner I talk to is using AI for something now. Very few of them can point to a decision it helped them make.
The numbers back that up. In a Goldman Sachs survey of small business owners this spring, 76% said they currently use AI, but only 14% said it was fully embedded in their core operations. Most owners are using it to draft emails and social posts. That’s useful, and it saves time. It just isn’t where the money is.
The usual advice for closing that gap is more training and better tools. Good advice. Necessary advice. But it skips over the reason most owners stall.
The Gap Nobody Talks About
Try asking an AI tool a question that actually matters to your business. Which customers cost you more to serve than they pay? Why did gross margin drop in August? Two kinds of answers usually come back. One is generic enough to apply to any company. The other sounds specific and confident, and it’s built entirely on whatever you pasted into the chat window.
I wouldn’t bet payroll on either one.
The trouble starts with where your information lives. A 25-person company might keep customers in a CRM, jobs in field service software, invoices in QuickBooks, payroll somewhere else, and the real pricing in a spreadsheet only one person understands. Intuit’s latest small business survey found that more than a quarter of small businesses run six or more digital systems, and the top challenge owners reported with their digital tools was the lack of integration between them (36%). More than half still use spreadsheets to manage their finances.
So, the AI walks into a business where the numbers don’t agree with each other. Your CRM says a customer spent $40,000 last year. Your accounting system says $31,000 because two invoices were coded to a different account. The AI has no way of knowing which one is right, and it won’t tell you it doesn’t know.
Salesforce reached a similar conclusion in its 2026 research: messy data is the number one blocker keeping small businesses from getting real value out of AI.
Your AI Is a New Hire With No Onboarding
When I explain this to owners, I use a hiring example.
Say you hire the sharpest operations manager you’ve ever interviewed. On their first morning, you hand over seven logins and a shared drive full of files named “Q3 FINAL v2,” with no explanation of how any of it connects. Then you ask which of your service lines actually makes money.
They’ll come back with an answer, and it might even sound right. But nobody showed them how a job in your field software ties to an invoice in your books, or how that invoice ties back to a customer in the CRM. They’d be guessing. A smart person guessing is still guessing.
That is where most small businesses are with AI right now. The tool is capable. It was never given the context.
Owners seem to sense this. In Bluevine’s 2026 small business AI report, 31% of owners named distrust of AI’s accuracy as a barrier to using it more, and 78% said they don’t fully trust it to handle even low-level tasks without someone checking. That instinct is reasonable. In a separate survey this summer, Databox found that 74% of business users had acted on an AI-generated number that later turned out to be wrong.
What Changes When Your Systems Are Connected
Most owners already pay for plenty of software. The work is getting those systems to share what they know, so the connections between them are clear: this customer, these jobs, these invoices, this margin.
Once that’s in place, a question like “Which of my top 20 customers haven’t ordered in 60 days?” takes a couple of minutes instead of an afternoon of exports and VLOOKUPs.
The bigger change is that answers come with receipts. When a number can be traced back to the record it came from, you can check it before you act on it, and your team stops arguing about whose spreadsheet is right. (If you want to see what that looks like in practice, here’s a plain-English walkthrough of tracing a number back to its source.) An answer you can’t check is really just an opinion, and owners already get plenty of those.
It also frees up whoever has been playing human integration layer, copying numbers between systems every Friday afternoon. In a lot of small companies, that person is the owner.
5 Steps to Start This Month
None of this requires a data team or a six-month project. It starts with knowing where your information actually lives.
- Write down every system that holds a number you care about. Include the spreadsheets, and include the one your office manager won’t let anyone touch.
- Pick the handful of questions you ask most often, the ones you’d love answered on a Monday morning without calling anybody, and write them in plain English.
- For each question, note which systems it pulls from. The questions that touch three or four systems are where your manual work and your conflicting numbers come from.
- Agree on definitions before you buy anything. What counts as revenue? An active customer? A profitable job? If your own systems disagree on those, no AI tool will give you a consistent answer.
- Before you act on any AI answer, ask where the number came from. If the tool can’t show you, keep it away from real decisions for now.
An owner who works through that list over a week or two will understand their data better than plenty of companies ten times their size.
Where I’d Start
Most of the conversation about AI in small business is about which product to buy. I’d start somewhere else: Can your business answer its own questions today with the systems it already has? If the honest answer is “only after someone spends a day in spreadsheets,” putting AI on top won’t fix that. You’ll just get wrong answers faster.
Connect what you have, settle on what your numbers mean, and then give the AI something real to work with. That’s the order that pays off.
Chris Rafter is Chief AI Officer of SIGNLD, a decision intelligence platform that connects the systems a business already runs so owners can get answers traced back to the source. Connect with him on LinkedIn, and follow SIGNLD on X at @signldai.
Photo courtesy Pablo Merchán Montes for Unsplash+
