Apollo's economist sees no AI payoff outside tech. What it means for your company

What Apollo said

On Oct. 24, Torsten Slok, chief economist at the investment firm Apollo, published a short note called No Signs of AI Payoff Outside the Tech Sector. His claim is blunt: AI has delivered "zero financial payoff to corporate America outside of tech."

He looks at what he calls the S&P 493: the S&P 500 minus the seven giant tech companies. According to his charts, revenue per worker and profit margins are rising for those seven and flat for the other 493. He also sees no sign yet of AI lifting productivity growth. He says mainstream companies are starting to see spending on AI models as a cost center that weighs on margins, not a source of growth. TechCrunch picked it up on Oct. 25, and it spread from there.

Slok also argues it may stay hard. In his view, business decisions run on judgment and context, and AI doesn't understand accountability, trust or unwritten rules.

Why it matters to a mid-size company

You aren't in the S&P 493. But you're being sold the same tools: chat assistants, copilots, AI add-ons to your ERP and email. If the biggest non-tech companies, with big budgets and IT staff, can't see a payoff in their numbers yet, a 150-person distributor should ask what it's actually getting.

The smaller-company data points the same way. The New York Fed's regional business surveys found 51% of manufacturers in its region used AI in 2026, up from 16% in 2024. But among firms using it, the median share of workers using AI was 7% at manufacturers and 17% at service firms. Most companies have AI somewhere. Few have changed how the work gets done.

Using AI is not the same as working differently

New York Fed regional business surveys, August 2026. Share of firms using AI, and the median share of workers using it at those firms.

1. Service firmsFirms using AI61%
2. Service firmsMedian share of workers using it17%
3. ManufacturersFirms using AI51%
4. ManufacturersMedian share of workers using it7%

Source: Federal Reserve Bank of New York, Liberty Street Economics, Sept. 2026. New York and northern New Jersey firms.

A license on 10 desks doesn't change a company's margins. A process that used to take 12 minutes per order and now takes three might. That's the gap Slok is pointing at, from the top down.

AI pays when it changes one process you can measure. A license on every desk is not a process.

Where the other side has a point

Slok's note is short and it's an economist's view of the biggest public companies, in aggregate. There are fair reasons to read it with care.

• It's early, and big averages move slowly. A gain in one department at one company won't move the margins of 493 giant firms. Slok himself says the economy-wide effects will stay very limited while heavy AI use is confined to a small part of the economy.
• Some companies do report value. BCG's Applied AI Index 2026, a survey of 1,330 executives, found 7.5% of companies it calls "future-built" plus a further 41% that are scaling AI and outperforming. Survey answers aren't audited results, but they aren't nothing.
• Revenue per worker isn't the only measure. Fewer errors, faster quotes and fewer late orders may show up as kept customers before they show up as margin.

Still, Slok has a point that's hard to argue with. If AI spending can't be traced to a result, it's a cost. BCG says the same thing in a friendlier way: the companies getting value reshape workflows end to end and measure the impact directly in their P&L.

Why measured beats felt

One reason the payoff is hard to see: people feel faster with AI even when they aren't. In a 2025 METR trial, experienced developers using AI tools were 19% slower but believed they were 20% faster. That was software developers, not an order desk, but the lesson carries over. If your evidence that AI is working is that people like it, you don't know yet.

The same pattern showed up in MIT's 2025 GenAI Divide report, which found about 95% of enterprise generative AI pilots had no measurable profit-and-loss impact, as reported by Fortune. We wrote about why in why AI pilots fail.

What to do this quarter

1. List what you pay for. Every AI subscription, add-on and seat. Next to each, write the process it changed. If the answer is "nothing specific," that's your first cut or your first fix.
2. Pick one process with units. Orders, invoices, POs, quotes. Something you can count.
3. Measure before and after. Minutes per item, items per person per day, errors per hundred. Our guide on measuring office automation ROI shows how to do it in a week.
4. Change the process, not just the tool. AI that reads a PO and fills in a draft order for a person to approve changes the work. A chat window next to the same old screen often doesn't.
5. Keep judgment with people. Slok's point about judgment is right for pricing exceptions, credit holds and customer disputes. Automate the reading and typing; leave the calls to your team.

If Slok is right, most companies will spend another year paying for AI that doesn't show up in their results. A mid-size company can be one of the ones that can prove it does, one measured process at a time.