What happened
On 25 October Torsten Slok, chief economist at Apollo Global Management, published a note that has been widely shared. His thesis fits in one line: outside tech, AI has delivered no financial payoff to US companies.
Slok compares two groups. On one side, the seven big tech companies: Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla. On the other, the remaining 493 companies in the S&P 500. In the first group, revenue per worker and margins are rising. In the second, they are flat. And he adds that many companies are starting to see AI spending as a cost that weighs on margins, not as a source of growth.
Not everyone agrees. An analysis by the Federal Reserve Bank of New York finds improvements in revenue per worker and margins since 2019, although it doesn't separate the big tech companies from the rest, so the two readings aren't fully comparable.
And in Spain, what does the data say?
Ten days earlier, on 15 October, Holded published its Emprende 2026 report, based on 1,530 interviews with entrepreneurs and self-employed workers across Spain. The figures paint a similar picture seen from below:
- 31% of small businesses already use AI, and 75% of entrepreneurs consider it key.
- Those who use it say they save 13.3 hours a week on average.
- The main use is creating content and marketing (47.8%), followed by data analysis (31.7%) and management and administration (30.65%).
- 42% notice higher productivity, but only 31% mention cost savings.
- Cost is the top obstacle for those already using it (37%), ahead of lack of knowledge (28.6%).
Why hours saved aren't money
This, for us, is the key to both stories. Saving thirteen hours isn't making money. It's freeing up thirteen hours. If those hours go into other tasks that don't sell more or cost less, the income statement doesn't notice.
And look at the main use: writing content. It's the easiest place to start, because all it takes is opening ChatGPT. But it's also the hardest to measure. Nobody knows how many sales a post written in ten minutes instead of an hour brings in.
What we see in the companies we work with is that AI makes money when it's applied to a specific process with a number behind it. Replying to a lead in two minutes instead of two hours. Sorting and answering customer service emails. Pulling the data out of supplier invoices without typing it in. That's where you can measure the before and after.
What it means for a small business
The good news is that a small business has it easier than a large S&P 493 company. It doesn't have to move thousands of people or go through five committees. It can pick one process, automate it and see the result within a month.
And the cost of the technology is no longer the excuse. As we explained in what automating with AI costs today, small models cost cents per task. The real cost lies in designing the process well, integrating it with your tools and maintaining it. If that isn't done, AI spending stays exactly that: spending.
What I would do this week
First, list what you already use. ChatGPT, Copilot or Gemini licences, tools with AI built in. Write down how much you pay each month for each one.
Second, for each use, ask which number it moves. Sales, response time, hours in a specific role, errors. If you can't answer, it probably isn't making money.
Third, pick a single process that does have a number. Ideally something repetitive, high-volume and tied to sales or costs. Measure how it works today before touching anything.
Fourth, automate it and measure for a month. If the number improves, expand. If not, change process. Our guide on when to implement AI and when not to sets out the criteria we use to choose.
And a personal opinion: if your team only uses AI to write emails and posts, the problem is rarely the tool. What's usually missing is training to use it on the real work.
If you want us to look at it with you
We design AI agents and automations measured by what they save or sell, we provide team training so AI gets out of the chat window, and as a fractional CTO we help decide where to invest and where not to.
Book a thirty-minute call. We go through what your company uses AI for today, what it's costing you and what the first process with a measurable return would be. You leave with a concrete proposal, or with the advice not to spend yet if it doesn't pay off. No commitment.
Sources
This article draws on the following publications, consulted on 26 October 2026:
Frequently asked questions
Torsten Slok, chief economist at Apollo Global Management, stated on 25 October 2026 that artificial intelligence has delivered no financial payoff to US companies outside the technology sector. According to his note, revenue per worker and margins are rising at the seven big tech companies but remain flat at the other 493 companies in the S&P 500.
It's a way of referring to the S&P 500 index without the seven big tech companies: Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla. It's used to see what is happening in the rest of the economy, without the weight of those seven companies.
According to Holded's Emprende 2026 report, based on 1,530 interviews with entrepreneurs and self-employed workers, 31% of Spanish small businesses already use AI. Those who use it say they save an average of 13.3 hours a week, mostly on content creation and marketing.
Because saving time isn't the same as making money. If the hours freed up aren't spent on selling more or cutting costs, the income statement doesn't change. On top of that, the most common use, writing content, is the hardest to measure. AI usually pays off when it's applied to specific processes tied to sales or costs and the before and after are measured.
With a single repetitive, high-volume process tied to a number: lead response time, customer service or invoice data entry, for example. It's worth measuring how it works before automating it, applying AI and comparing the result for a month before expanding.
The technology is no longer the expensive part: small models cost cents per task. What costs money is designing the process well, integrating it with the tools the company already uses and maintaining it. Even so, cost is the top obstacle cited by small businesses already using AI, according to Holded's report.