On Saturday, October 24, Apollo Global Management chief economist Torsten Slok published a short note arguing that AI has produced no measurable financial return for large US companies outside the biggest technology firms. Within two days SiliconANGLE, TechCrunch and PYMNTS had each written it up, and each read it a little differently. Taken together, the coverage says more about AI spending in the sales stack than any one of the stories does.
What Apollo published
The note, titled No Signs of AI Payoff Outside the Tech Sector, compares the seven largest US technology companies with the rest of the S&P 500, which Apollo calls the S&P 493. It uses five charts. Revenue per worker and profit margins are rising for the seven and flat for the 493, and a productivity series from the San Francisco Fed shows no AI lift. Two more charts, sourced to Ramp, track token spending, and the note says venture-backed companies are spending more on tokens while mainstream corporate America does not need the latest model and is unwilling to pay for it.
The opening line is the one every outlet picked up: “Despite massive investments and widespread hype around artificial intelligence, the technology has delivered zero financial payoff to corporate America outside of tech.” The note goes on to say that “mainstream corporate America is pulling back and starting to view token spending not as a source of growth but as a cost center that weighs on margins.” Slok closes by arguing that the AI story will be decided in the S&P 493, and that so far there is no evidence of better revenue growth or margins outside tech.
Advertisement
Four days earlier, on October 20, Slok published a related note, Slowing Token Spend Shows Compute Demand Isn’t Unlimited. It describes corporate demand for compute as highly price-sensitive and says “even modest changes in token prices can drive buyers to throttle usage or substitute cheap open-source alternatives.”
How three outlets told it
SiliconANGLE: the spending pullback
SiliconANGLE’s account stayed closest to the charts. It read the revenue-per-employee chart as rising from an average of $550,000 in 2022 to almost $800,000 this year at the seven tech firms, against a line that has sat near $550,000 since 2019 for the rest. It gave the token-spending passage a full paragraph, including Apollo’s point that venture-backed startups are spending more on tokens while established firms hold back. It then walked through Slok’s argument about judgment, context and accountability.
TechCrunch: the limits of machine judgment
TechCrunch ran it as a brief and disclosed that Apollo used to own TechCrunch. After the headline finding, it spent most of its space on the passage about judgment, quoting Slok’s list of what AI does not understand: accountability, trust, culture, unwritten rules, ethical trade-offs and fairness. It added that recent studies have also found no significant productivity gains from AI so far. It did not mention token spending.
PYMNTS: earnings season and adoption data
PYMNTS placed the note in earnings season. It said Bloomberg News flagged the note on Sunday, October 25, and set it beside Meta’s third quarter, in which Meta reported $34.9 billion of capital expenditure and negative free cash flow of $2.1 billion. It then added its own survey data. According to PYMNTS Intelligence, 7% of executives it surveyed in 2025 were using or piloting AI agents, while in 2026 89% use agents in at least one part of the business. It reports that 78% would let agents read company systems but not act on them, 18% would let them act with a human’s approval, and 2% would let them run routine work alone.
Where the accounts differ
The three agree on the headline and on Apollo’s figures. They split on cause. SiliconANGLE and PYMNTS put the weight on money: firms treating token spend as a cost, and investors watching capital spending as results arrive. TechCrunch put the weight on capability, presenting the problem as AI’s weak grasp of business judgment. Those readings point to different fixes. If the problem is cost, cheaper models and tighter usage control help. If it is judgment, cheaper models change little.
PYMNTS also brought the one piece of evidence that sits between them. Its survey describes companies that have let agents in and then kept most of them read-only. An agent that can read a system and cannot act on it is a cost with a narrow route to revenue, which fits both the cost reading and the judgment reading at once.
What the coverage leaves out
None of the three accounts asks which kinds of AI spending inside a company can show a return, and Apollo’s note does not either, because it measures whole companies. Revenue per worker and profit margin are company-wide numbers. A sales organisation is one of the few places inside the S&P 493 where AI spend can be tied to revenue at the level of a team, a rep or a deal stage, because the CRM already records who touched which opportunity and what it closed for.
In our reading, that makes sales the place where Apollo’s cost-center view will be tested first. When a finance team starts treating token spend as overhead, the AI line items that survive will be the ones a revenue leader can connect to pipeline created, meetings booked or deals closed. AI line items with no such link will be easier to cut.
A vendor-side data point
Two days before Apollo’s note, Podium, which sells AI agents to local businesses such as car dealers, HVAC contractors and aesthetics clinics, said it had passed $150 million in AI-specific annual recurring revenue in two and a half years. Its agents answer calls, respond to leads, book appointments and follow up with customers. Podium says more than 20,000 of its agents are deployed and that they have helped customers generate more than $13 billion in revenue, a company figure that Podium did not break down. Its CEO, Eric Rea, described the bet this way: “In the age of AI, you can either disrupt your own company or wait for somebody else to do it.”
Podium sells to local businesses, so the release says nothing directly about Apollo’s sample of the largest US companies. What it shows is the shape of AI spending that buyers keep paying for: agents assigned to a revenue task with an output a business owner can count, such as a booked appointment or a lead answered.
Pricing will decide part of it
Apollo’s October 20 note matters here as much as the October 24 one. If corporate buyers cut usage or switch to open-source models when token prices move, then sales AI priced by consumption carries budget risk for the buyer and revenue risk for the vendor. We have argued before that teams should stop signing token-metered AI sales-tech contracts without caps, and that AI spend risk is moving from the contract to the control panel. Apollo’s two notes give finance teams a macro reason to ask the same questions.
What it means for the sales leader
Expect the Apollo framing to reach budget reviews. A CFO who reads that mainstream companies see token spend as a cost center will ask the CRO to show which AI tools move revenue. A sales leader who can answer with the team’s own numbers is in a stronger position to keep the budget.
Four steps help before that conversation happens:
- Measure revenue per rep before and after. Apollo’s first chart uses revenue per worker. Run the same measure for each team that adopted an AI tool, with a comparison team or period that did not.
- Tie each AI line item to a stage. List every AI product in the sales stack against the pipeline stage it is meant to change, and the CRM field that shows the change.
- Separate read-only tools from tools that act. The PYMNTS survey suggests most agents still only read. Tools that only summarise or recommend need a clear time-saved figure; tools that act need an outcome figure.
- Check how each contract prices usage. Find which tools bill by tokens, credits or actions, what the cap is, and what happens to cost if usage doubles.
Apollo’s charts will be updated, and the next earnings season will add more data. Sales teams that set up the measurement now will have a year of their own figures by the time the question comes back.
Source: Apollo, The Daily Spark: No Signs of AI Payoff Outside the Tech Sector
