The Signal #116 — Dakota’s read on the AI news that actually matters to people running a business.

Everyone has been waiting for the AI productivity boom to show up in the numbers. According to Apollo’s chief economist, it has not. Not outside of tech, anyway.

This is worth reading carefully, because the conclusion has real implications for how you think about AI spending right now.

What happened

Apollo chief economist Torsten Slok published a piece on October 24th titled No Signs of AI Payoff Outside the Tech Sector. The core finding is blunt: despite massive investment and widespread adoption, AI has delivered zero financial payoff to corporate America outside of tech.

The data Slok presents breaks down along a clear line. Revenue per worker is rising for the Mag 7 (the seven largest technology companies by market cap). It is flat for the S&P 493 (meaning the other 493 companies in the S&P 500, which represent a wide cross-section of mainstream corporate America). Profit margins tell the same story. Rising for the Mag 7, flat for the S&P 493. And total factor productivity (a broad economic measure of how efficiently an economy turns inputs like labor and capital into output) shows no signs of improvement either.

Slok also points to a behavioral shift happening in response. Mainstream corporate America is starting to treat token spending (the cost of sending text to an AI model and receiving a response, billed per unit of text processed) not as a growth investment but as a cost center that weighs on margins. VC-backed companies are spending more on tokens. Mainstream operators are pulling back.

The closing observation is the one that should sit with you. Slok writes that AI may be a great calculator, but in business, intelligence is mostly judgment. And judgment depends on context. AI does not understand accountability, trust, culture, unwritten rules, ethical trade-offs, or fairness, all of which shape reputational risk and ultimately drive revenue growth and margins.

Why it matters

If you are running a real estate brokerage, a regional accounting firm, a mid-market e-commerce operation, or a healthcare administration group, this data is basically about you. You are in the S&P 493 category of the economy. And the numbers say that, so far, your peers have not figured out how to turn AI spending into measurable revenue or margin improvement.

That cuts two ways.

First, it is an honest reality check. A lot of operators have been told that AI adoption is urgent and that the productivity gains are obvious. The macro data does not support that story yet. Slok’s framing is useful here: the question looking ahead is whether mainstream corporate America can turn increasingly cheap intelligence into higher revenue growth and higher profit margins, and he adds that may be harder than it sounds.

Second, it identifies where the actual work is. The companies getting financial returns from AI right now are the ones closest to the model infrastructure itself. The rest of the economy is still in the phase of spending on AI without having figured out how to route that spending toward outcomes that show up on the income statement.

That gap is not permanent. But it does mean the default path of buying a tool and hoping it moves the needle is not working for most operators.

What most people get wrong

The wrong read on this report is that AI does not work, full stop. That is not what the data says.

What it says is that diffuse, unfocused AI adoption does not produce measurable financial results. Token spending (the cost line that shows up when your team uses AI tools) becomes pure overhead if it is not connected to a specific business output. Lower cost per output unit. Faster cycle time on a billable process. Fewer errors in a workflow that currently requires rework. Something concrete.

The mistake most operators are making right now is treating AI as a capability upgrade rather than a process redesign. You do not get margin improvement by giving people a faster way to do the same thing in roughly the same amount of time with roughly the same error rate. You get it by identifying a specific process, understanding where the time and cost actually go, and building AI into that process in a way that measurably changes the output.

Slok’s observation about judgment is also worth taking seriously. The places where AI struggles, context, accountability, trust, unwritten rules, are exactly the places where human operators add disproportionate value. Knowing the difference between those two categories in your own business is more important than picking the right model.

The actual lesson

The AI story will be decided in the S&P 493. That is a direct quote from the report, and it is the right frame. The technology companies are already capturing gains. Everyone else is in a prove-it phase.

Proving it means being specific. Which process, which cost, which output, which number changes if this works. Operators who can answer that question clearly are the ones who will eventually show up on the right side of this data. Operators who cannot will keep adding to the cost line without moving the revenue or margin line.

If you want help thinking through where AI actually fits in your specific operation, start at xovionlabs.com.