Short answer: AI can pay off, but only for companies that write down what they expect from it and then actually measure it. This week two serious analyses gave almost opposite answers to the question of whether AI makes companies money. One says no. The other says almost half of companies already see value. Both can be true. And that is the real lesson for a small or mid-sized business.
What happened?
On 25 October TechCrunch reported on a note by the chief economist of Apollo Global Management. The economist looked at the S&P 500 without its seven largest tech companies, a group he calls the "S&P 493". For those companies, profit margins and revenue per worker have been essentially flat, and he sees no evidence that AI is lifting productivity. His conclusion is blunt: despite massive investment and hype, AI has delivered "zero financial payoff" to corporate America outside tech. In his view AI may be a great calculator, but most business decisions are about judgment, accountability and trust.
A few weeks earlier, on 30 September, BCG published its Applied AI Index 2026, based on a survey of more than 1,300 senior executives. It paints a different picture. According to BCG, almost half of companies now generate value from AI. The most mature, "future-built" companies, 7.5% of the sample, show 2.4 times the top-line growth of the companies in the bottom half. Corporate AI spending doubled in a year to 3.3% of revenue, and 80% of it now sits outside the IT budget.
The two pictures don't contradict each other. Apollo looks at the average across large listed companies. BCG shows who is winning inside that average, and what they do differently.
What does this mean for an SMB?
The most important figure isn't the 2.4x growth. It is that, according to BCG, 95% of the most mature companies track AI value with clear KPIs or directly in their P&L. Hungarian trade magazine Computer Trends, summarising the report, adds that these leaders follow a 10-20-70 rule: 10% of resources go to algorithms, 20% to technology and data, and 70% to people, processes and change management.
For a smaller company, three things follow.
"We use it" is not the same as "it pays". A typical situation: a company subscribes to an AI assistant (Claude, ChatGPT, Copilot or Gemini), people like using it, and six months later nobody can say what it delivered. Not because it delivered nothing. Because nobody recorded how long the task took before.
The average hides the spread. If the average isn't moving, as Apollo says, while part of the market is clearly gaining, as BCG says, then plenty of companies are paying for AI and getting little back. For an SMB that is not an argument against AI. It is an argument against "rolling out AI" in general, and for using it to fix one specific process.
Most of the money doesn't go on software. The 10-20-70 split comes from large enterprises, but the logic holds at any size. The tool's price is the smaller item. The bigger one is whether you redesign the process, train the team, and make someone accountable for the result.
There is also a timing point. The Apollo argument is about profit margins, which move slowly and reflect thousands of decisions at once. A single automated process in a small company can show its effect within a quarter, if you know what you are looking for. Waiting for the macro data to settle the debate is not a plan.
What should you do now?
- Write down in one sentence what your next AI or automation project must replace: which process, how much time or how many errors. If you can't write it down, it isn't ready to start. Our article on what to automate first helps you pick the process.
- Measure the starting point before go-live: time per item, errors per month, current cost. You can't reliably reconstruct this afterwards.
- Count the full cost, not just the licence: implementation, training and the time people spend checking the output.
- Schedule a follow-up measurement at three and six months, using the same metrics. If the number hasn't improved, stop or redesign the project instead of quietly renewing the subscription.
- Name one owner who is accountable for the result, not for the tool. We collected the detailed steps and a one-page template in our guide on how to measure the ROI of an AI or automation project.
If you want your next automation project to start with a measurable goal, and to find out where it really pays off, take a look at our process automation service.
