The AI Payoff Problem: 5 Reasons Most Firms See Little Return

The AI payoff everyone promised has not shown up for most companies outside big tech, according to a blunt new note from an Apollo Global Management economist.

TechCrunch reported on October 25, 2026 that Torsten Slok, Apollo’s chief economist, looked at the “S&P 493”: the S&P 500 minus Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla. What he found was uncomfortable. Profit margins and revenue per worker look flat. No visible productivity lift. His summary: AI “has delivered zero financial payoff to corporate America outside of tech.”

What the Apollo Note Actually Says

Slok does not buy the theory that everyone else is simply late to the party. In his view, it will be hard for businesses to “turn increasingly cheap intelligence into higher revenue growth and higher profit margins.”

Then comes the line that stuck with me: “AI may be a great calculator, but in business, intelligence is mostly judgment.” He argues that AI does not understand accountability, trust, culture, unwritten rules, ethical trade-offs or fairness. Those things drive reputational risk, and reputational risk drives revenue.

TechCrunch adds that other recent studies have also failed to find significant productivity gains so far. So this is not one economist having a bad day.

Reason 1: Tools Were Bought Before Problems Were Defined

Let me be direct. A lot of AI spending started with “we need an AI strategy” and not with “this process costs us 400 hours a month and we want to cut it.”

When the goal is vague, the result is vague. Teams roll out a chatbot, count the licenses, and call it transformation. Nobody can point to a line in the P&L that moved.

Reason 2: Cheap Intelligence Does Not Equal Better Judgment

Slok’s point about judgment deserves a second look. Drafting a report is cheap now. Deciding whether the report’s conclusion is safe to act on is not.

In regulated industries, a confident wrong answer costs far more than a slow right one. That is why human review stays in the loop, and why the time saved on drafting often gets spent on checking.

Reason 3: Old Workflows Wrapped in New Tech

Bolting AI onto a broken process gives you a faster broken process. The McKinsey State of AI research has repeatedly pointed to workflow redesign as one of the strongest drivers of value from generative AI. Companies that only add a tool on top rarely see the same results.

The fix is boring: map the process, remove steps that exist only out of habit, then decide where automation belongs. Skipping that mapping is how pilots end up as slide decks.

Reason 4: Data That Is Not Ready

Models are only as useful as the data they can reach. If customer records live in four systems that do not talk to each other, your shiny assistant will confidently summarize the wrong one.

Gartner has warned that many AI projects get abandoned when data quality and governance are not in place. You can read the broader Gartner AI research hub for their latest forecasts. The takeaway is consistent: data readiness is not glamorous, but it decides outcomes.

Reason 5: Nobody Owns the Result

Who is accountable if the AI project misses its target? In many firms the honest answer is “nobody.” IT ran the pilot, the business unit used it, and finance never saw a number.

Successful teams assign a business owner, set a baseline before launch, and review the metric monthly. Without that, you cannot tell a real gain from a good story.

What This Means for Pakistani and Global SMEs

If you run a small or mid-sized business, this note is useful, not discouraging. Big companies burn money on experiments you do not need to copy. You can start narrow: one workflow, one metric, one owner.

Good starting points include customer support triage, invoice processing, internal knowledge search and code review assistance. Pick the one where you already measure cost or time. Run it for 60 days. Compare against the baseline. Yeh seedha hisaab hai.

The OECD AI policy pages are also handy if you need a neutral reference on how governments and firms are thinking about productivity and responsible adoption.

A Simple Test Before You Spend

Ask four questions before approving any AI project. What process does it change? What number will move? Who owns that number? What happens if the model is wrong? If you cannot answer all four in two sentences each, the project is not ready.

It is a small filter, but it would have saved plenty of companies from the exact trap Slok describes.

Key Takeaways

• The gap is real: Apollo’s analysis of the S&P 493 shows flat margins and flat revenue per worker so far.
• Judgment is the bottleneck: AI is a strong calculator, but accountability and trust still need people.
• Redesign the workflow: Adding a tool to a weak process rarely changes the outcome.
• Fix data first: Fragmented, low-quality data limits what any model can deliver.
• Assign ownership: Set a baseline, name an owner and review a single metric every month.

How TecniForge Can Help

At TecniForge, we help businesses navigate these technology shifts. Whether you need custom software development, AI integration, or cloud migration, our team builds scalable solutions tied to measurable business outcomes. Talk to our experts.

One last thing to chew on: if you switched off every AI tool in your company tomorrow, which number on your dashboard would actually change?