The 2026 QuickBooks AI Impact Report included a survey of 34,000 small businesses located throughout the U.S. in partnership with the University of Chicago. On paper, the data looks impressive – an increase to 77% of respondents who reported that they used AI on at least a regular basis, up significantly from 48% in mid-2024. In addition, 41% of those surveyed reported increases in revenue due to their use of AI as well as an improvement in their productivity by 74%. Numbers like these reveal one side of the story about the use of AI for small business purposes. However, the “fine print” of this study reveals another side. For measuring AI usage within a small business setting, the most important issue is not whether a company uses AI but whether or not it has the capability to measure if its AI is successful.

When the researchers asked about how they (businesses) were measuring the improvements from these gains, the answers became quite vague. More than 50% stated a general feeling that their business was better. Less than half of respondents reported tracking specific metrics. The productivity number was based on self reporting and not based on time studies. The revenue attributed to “AI” was based on correlation as opposed to controlled measurements. This is also the area that most small businesses currently reside within. As such, Gartner predicts that by 2026 global AI spending will be at $2.52 Trillion. A large portion of this money has been spent with no real method to measure the return.

Why Most Businesses Cannot Measure AI Results

The AI has to begin at a point after that of the Measurement Problem. Many Small Businesses lack Baseline Data (the amount of time it takes humans to accomplish tasks) they are passing off to the AI. Without this data, there can be no way to measure what percentage of time the AI saves in accomplishing tasks. Likewise without tracking conversion rates of previous Human-Generated Campaigns, there will be no basis to compare to those generated by the AI. The Baseline is missing. Thus, there will be no measurable comparison of Improvement.

The second problem is how to assign credit (attribution) for changes when a business uses an artificial intelligence system on emails to customers and also hires a new sales person during the same time period. The business owner will be unable to determine if it was the use of the AI that resulted in increased revenue, the hiring of a new sales person, changes due to the seasons, or a change in the market that had no connection to technology. I previously discussed the AI investment gap. The fundamental issue remains the same. Companies are purchasing systems but there are no definitions of successful outcomes prior to beginning.

The 55% was a prediction by Forrester that many (55%) AI projects will not meet their intended goals. This does not necessarily mean that an initiative failed because of its technology, it simply means that there was no way to measure success at the time of the project’s launch. A successful tool can appear unsuccessful when a person has never clearly established how “success” is defined prior to its use. Similarly, a project may seem successful with regards to its metrics but add little or nothing to the organization as long as the metrics are broad and inclusive of chance events.

Five Numbers Every Small Business Should Track

The five ways to measure AI ROI I published last year still hold up. But the QuickBooks data shows that many businesses are not tracking them. Here is a simplified AI measurement framework any small business can start this week.

Five Numbers Every Small Business Should Track

Source: Institute Of Business AI

First, we have to determine how much time each task requires. Choose three of your most used AI-based tasks. Record the time (in seconds) required to complete those same tasks before using AI. Use an actual stop watch or timer to record the times. Then compare the two times and multiply the time difference by your hourly salary to calculate your dollar amount of time saved. For example if a previously 1.5 hours task now requires only 0.33 hours using AI then this represents 1.17 hours saved. With an hourly wage of $100 for the business owner’s time this represents $116 in time savings per occurrence. Record these numbers on a weekly basis.

Second, output quality. Track how much editing (if at all) is required for each piece of content generated by an artificial intelligence. If 80% of your AI created content requires only minor adjustments then you are getting value from the AI tool. If 80% of your AI created content needs a full rewrite as well as the time to do so (with no net time savings), then you have simply changed your workflow with no additional productivity.

Third, the revenue per AI-supported activity. Compare revenue from campaigns written using AI with those drafted by humans for the same number of customers and during the same period of time. The two sets should have comparable customer lists, offer types, etc. The amount of revenue that you generate with an AI-drafted campaign compared to a human-drafted campaign will be your AI marketing ROI. If there are no differences in revenue, then the use of AI saved you time drafting email marketing campaigns however, this is an entirely different value than generating increased revenue.

Fourth, error rate. When using AI for tasks such as processing data entry, bookkeeping or report generation monitor how often you have to correct errors. Compare your current error rate to what it was when you did those tasks manually. In the event that there are fewer errors now than previously, then your AI is likely increasing the level of accurate work. Conversely, if the number of errors remains the same or increases and you’ve also increased your overall workload, then the AI has introduced additional errors requiring greater amounts of time for review.

Fifth, tool cost vs Value delivered. Calculate how much money all of the AI subscriptions your company has are costing. Next, divide that number by the documented value (from the previous measurements) that those subscriptions produce. For example, if your company spends $500 each month on AI and you’ve measured an additional $2000 each month as a result of using these tools (time saved and/or increased income), then you clearly see the ROI. However, if your company spends $500 each month and you can’t identify at least a certain amount of dollars that it produced in return, you either need to fix how you measure or what tools you use, likely both.

The Danger Of Adoption Without Measurement

The U.S. Chamber of Commerce has reported that 58% of all small businesses are now using generative artificial intelligence (AI), while an impressive 93% anticipate their business will grow in 2026. That is great news for business optimism. However, if you can’t measure your results with some degree of certainty or proof then you will continue to spend money on unproven tools. Even worse you’ll be tempted to buy even more tools as each new tool claims to fix problems that possibly were never addressed by previous tools.

The “trust” McKinsey’s 2026 AI Trust Survey shows companies to be able to invest in the building of a trust infrastructure (including measurement and governance) for their AI projects will build them faster as well as reduce the number of compliance problems they have. This idea holds true in smaller businesses. Small businesses’ teams can develop “trust” with the AI by showing them how much time it has saved, or how much money it has made.

In addition to restaurants, other businesses report similar results. For example, a restaurateur said that after using an AI-based reservation service, their bookings rose by 20%. He attributed this rise in bookings by comparing his bookings for May of this year versus May of last year. However, when booking comparisons are made without accounting for differences such as changes in customer review activity on websites like Google, number of dining areas (outdoor vs. indoor), or even menus offered at the time of comparison, then the data collected does not isolate variables but instead becomes anecdotal stories rather than evidence based data.

Small businesses do not have to hire a Data Science Team for proper measurement of Artificial Intelligence. What they need is discipline. Determine what metrics are being measured When will those metrics be started? Determine if there were changes made at the exact same time as launching the AI email campaign. There was probably hiring of a new Sales person that occurred during the same month. Measure each separately. Track revenue from deals generated by the sales person in one column, track revenue generated from AI campaign in another. Separation provides clarity-not bundling every single metric into one group and hoping that the overall picture looks good.

The 77% adoption rate by QuickBooks is true. Many of these companies will realize a gain in productivity from AI. However, “many” is not a business plan. Businesses that will be able to take advantage of AI during the next 12 months will be those that stop making educated guesses (i.e., “I think this will help”) and start counting. Identify three things that need to be automated. Record how much time they use. Rate their quality compared to before. Calculate what percentage of their total revenue comes from the work being done by AI. Calculate the costs associated with using AI. These numbers will show you if the AI is working or not. The same numbers will also identify where to correct/fix the problem(s).

Run the measurement for 30 days. If you receive good results then you will have the evidence needed to make the decision to further invest in this area. If you do not see improvement, that is a reason to evaluate how you can improve your strategy before wasting one more dollar. In either case you will know. Most businesses currently operate on guesses. Be different than most.

Source link