34,000 Small Businesses Say AI Is Working. Most Can't Prove It.

I talked to a client last week who swore AI had doubled his team's output. I asked him how he knew. "It just feels faster," he said. "Everyone says so."

That gut feeling is running most of the AI playbook right now.

QuickBooks just released its 2026 AI Impact Report, built from a survey of 34,000 U.S. small businesses in partnership with the University of Chicago. The headline numbers look great. 77% of respondents now use AI on a regular basis, up from 48% in mid-2024. 41% said AI increased their revenue. 74% reported productivity gains.

But the researchers dug deeper. They asked those businesses how they were actually measuring these gains. The answers fell apart.

More than half said it was just a feeling. Not a metric. Not a before-and-after comparison. A feeling that things were running better. Less than half tracked any specific numbers at all. The productivity stat was self-reported. The revenue figure was correlation, not controlled measurement.

Forrester predicted that 55% of AI projects won't meet their intended goals. Not because the technology failed, but because nobody defined what success looked like before they started.

This tracks with what I see every week consulting with small businesses in Austin. Someone signs up for three AI tools at $50 to $200 a month each. They use them for a while. Things seem fine. But when I ask them to show me the numbers, they can't.

The Baseline Problem

If you don't know how long a task took before AI, you can't prove AI made it faster. If you don't track conversion rates from your old email campaigns, you can't say the AI-written ones performed better. You're comparing to nothing.

The attribution problem is just as tricky. Say you launched an AI email campaign the same month you hired a new salesperson. Revenue went up 15%. Was it the AI? The new hire? Seasonal demand? Without separating those variables, you're guessing.

Gartner puts global AI spending at $2.52 trillion for 2026. A big chunk of that money is being spent with no way to prove what it's producing.

Four Things You Can Measure This Week

Pick three tasks you're using AI for right now. Time them with a stopwatch, both the old way and the AI way. If your AI drafts took 20 minutes where you used to spend 90, that's 70 minutes saved. Multiply by your hourly rate. Now you have a real number instead of a feeling.

Track edit rates on AI content. If 80% of what your AI produces needs only light editing, it's adding value. If 80% requires a full rewrite, you've just rearranged your workflow without saving any time.

Run one A/B comparison on a single marketing campaign. Send an AI-drafted version to half your list and a human-drafted version to the other half. Same audience size, same offer. Compare the revenue. That number tells you more than a month of gut feelings ever could.

Calculate your total AI tool spend per month. Then set it next to the documented savings and revenue those tools produced. If you're spending $500 and you can point to $2,000 in measurable value, keep going. If you can't point to anything specific, that's your signal to fix the measurement or fix the tools.

Optimism Is Not a Business Plan

The U.S. Chamber of Commerce reports 58% of small businesses now use generative AI. 93% expect to grow this year. Both numbers are encouraging. But optimism without evidence is just hope. Hope is fine for motivation. It's terrible for budgeting.

Run the measurement for 30 days. You don't need a data science team. You need a spreadsheet, a timer, and the discipline to write down the numbers before and after. If the results are good, you'll have proof to invest more. If they're not, you'll know exactly where to cut.

Most businesses are operating on educated guesses right now. The ones who start counting will pull ahead, because they'll know which tools actually earn their keep.

— Mark Garza, Laimen AI