Apollo's Slok: AI 'zero financial payoff' outside tech

Apollo Global Management's chief economist just said what a lot of operators have been thinking. In a new note covered by TechCrunch on October 25, Torsten Slok argues AI has delivered "zero financial payoff to corporate America outside of tech." The AI ROI question has moved from blog debates to a private equity research desk. That is worth paying attention to, and it is not a reason to stop building.

What actually happened

Slok looks at the "S&P 493": the S&P 500 minus Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla. Per TechCrunch, he finds profit margins and revenue per worker in that group flat, with no evidence yet that AI is lifting productivity. His line on why: "in business, intelligence is mostly judgment." He questions whether non-tech companies can turn cheap intelligence into higher revenue growth and margins.

This is not a new position for him. In a June 30 Daily Spark note, Slok wrote that "so far there are no signs of profit margins rising outside the tech sector." He flagged regulated, process-heavy sectors such as health care, banking, energy and manufacturing as the slowest to see gains, because the work has to be redesigned first. And he warned that markets priced for fast payoff face a repricing if the productivity curve takes "five years rather than five months."

TechCrunch's write-up does not reproduce specific figures from the new note, so we are not quoting any.

Why it matters for your business

Big-company averages hide where AI actually pays. A flat margin across 493 giant firms says little about whether a quoting bot saves your estimator ten hours a week. Aggregate data averages away the workflows that work and the pilots that never shipped.

Slok's point about redesign is the real lesson. Bolting a chatbot onto an unchanged process rarely moves a number. The wins we see come from picking one workflow, rebuilding it around what the model is good at, and keeping a human on the judgment calls.

Expect vendors to feel this. If public markets start pricing AI on delivered returns instead of promised ones, your vendors' pricing and roadmaps shift with them. Contracts you can exit, and systems you own, age better than bets on one platform.

Measure before and after, or you are guessing. Hours per task, error rate, cycle time, cost per run. Write the baseline down before the build starts.

Sources: TechCrunch, Apollo Daily Spark.