Let me say something that might be uncomfortable to hear in mid-2026: most small business owners have no idea if their AI tools are working.
Not because they're doing something wrong. Because they never set up a way to find out.
The 2026 QuickBooks AI Impact Report - built on responses from more than 34,000 small and midsize businesses across the US, Canada, the UK, and Australia, with analysis from economists at the University of Chicago - looks impressive at a headline level. Usage is up sharply. 77% of respondents now use AI at least regularly, compared to 48% in mid-2024. 41% say their revenue has increased. 74% report productivity improvement.
But read the fine print.
When researchers asked how businesses were measuring those gains, the answers got vague. More than 50% described their improvement as "a general feeling that their business was better." Less than half were tracking any specific metrics at all. The productivity number came from self-reporting, not time studies. The revenue gains attributed to AI were based on correlation - not controlled comparisons.
In other words: most small businesses believe AI is working for them. They just can't show you why.
Why This Is Almost Universal
Forbes contributor Terdawn DeBoe - who analyzed the QuickBooks data in detail in May 2026 - identified two core problems.
The missing baseline. Before you adopted AI, how long did the task take? Most businesses never recorded that number. Without a before, you can't calculate a difference. You can only guess.
The attribution problem. If your revenue went up after you started using AI-generated email campaigns, but you also hired a new salesperson during the same period, you cannot tell what caused the change. The credit is unassignable. You're left with a feeling, not a fact.
These aren't exotic data-science problems. They're the reason Forrester predicts that 55% of AI projects in 2026 will not meet their intended goals. Not because AI doesn't work - but because nobody defined "working" before they started. A tool that performs fine still looks like a failure if you never agreed on what success looked like.
This is also why Gartner estimates $2.52 trillion will be spent globally on AI in 2026 - much of it with no measurement framework attached. It's not just small businesses. It's most businesses.
If you've been feeling vaguely uncertain about whether the $100 to $400 a month you're spending on AI tools is worth it - you're not behind the curve. You're in the majority.
The Real Cost of Measurement Failure
There's a second-order problem here that doesn't show up in the survey data.
If you can't tell what's working, you can't make good decisions about what to keep paying for. You end up carrying tools you don't use, canceling things that actually were delivering, and layering on new subscriptions because you hope the next one will be clearer.
The "AI tool sprawl" pattern that's emerging - small businesses with 5, 6, 8 AI subscriptions, overlapping in function, many barely touched - is what happens when nobody is measuring. You buy based on demos and promises because you don't have a framework for evaluating reality.
The fix is not complicated. But it does require doing something before the next subscription renewal.
Five Numbers to Start Tracking This Week
These come from the measurement framework outlined in DeBoe's analysis - adapted for a small business without a data team.
1. Task time before and after. Pick three tasks you use AI for regularly. Set a timer for the old way, then set a timer for the AI-assisted way. Write the numbers down. Do this for two weeks. You'll have actual data.
Example: if a task that used to take 90 minutes now takes 20 minutes with AI assistance, and you do it weekly, that's about 70 minutes saved per week. At your effective hourly rate, you can price that in real dollars.
2. Output quality - edit rate. For content AI generates, track how much editing you're doing. If you're making minor tweaks to 80% of the output, the tool is delivering value. If you're doing full rewrites - adding back your voice, fixing wrong information, restructuring the argument - you're not saving time. You've just moved the work.
3. Revenue per AI-assisted campaign vs. human-drafted. For marketing specifically: run the same type of campaign in parallel - one AI-drafted, one written by you or a team member, same offer, similar audience size. Compare the numbers. This is the only way to know if AI-generated copy actually converts for your business.
4. Error rate. If AI is handling data entry, bookkeeping tasks, or report generation, track how often you have to correct it. Compare to your error rate before. Fewer errors is a real gain. The same or more errors - with more total volume to review - means AI has added a QA burden you didn't have before.
5. Tool cost vs. documented value. Add up every AI subscription you pay for. Divide by the total documented value from the measurements above. If you're paying $400/month and can only document $200/month in measurable benefit, you know exactly what to cut.
DeBoe's framework adds one important qualifier: if you find a gap in your measurement, fix how you measure before assuming the tool is failing. A tool can work fine and look bad if you're measuring the wrong thing.
What to Do Right Now
You don't need a spreadsheet system or a dedicated analytics session to start. You need three things:
A timer. Use it before and after a task for one week. Write the numbers somewhere.
A simple log. A shared note or doc where you record when you use AI, for what, and whether you edited the output substantially or minimally.
A subscription audit. Go through your credit card statement and list every AI tool you're currently paying for. Next to each one, write what you last used it for and whether you have any data on whether it helped.
That's it. Not a framework project. Not a consulting engagement. Thirty minutes of honest accounting.
The QuickBooks data shows that most small businesses are operating on faith right now - not certainty. That's fine for the short term. But at some point, "I have a general feeling it's helping" is not a good reason to keep paying $300/month for a stack of tools.
The businesses that will use AI well in 2026 are not necessarily the ones that adopted the most tools. They're the ones that knew - specifically, in writing - what they were trying to accomplish and whether it happened.
Sources: QuickBooks 2026 AI Impact Report (34,000 businesses, University of Chicago collaboration); Forbes / Terdawn DeBoe, May 29, 2026; Forrester 2026 Predictions; Gartner AI Spending Forecast, January 2026