Here is the full Benderson Media article: —
AI Has a Zero Return Problem and Most Companies Are Hiding It
Companies are burning billions on AI and the books don’t show it. According to Apollo Global Management’s chief economist Torsten Slok, most businesses are getting ly zero financial payoff from their AI investments right now. Not low returns. Zero. And if you’re still waiting for the productivity miracle to show up in your revenue, you might be waiting a long time.
Why This Is Blowing Up Right Now
The timing matters. We’re in 2026 and the AI spending wave started in earnest in 2023. That’s three years of enterprise budgets redirected toward tools, talent, and infrastructure. By this point, the returns should be showing up somewhere measurable. They’re mostly not.
According to Goldman Sachs in their widely cited 2024 report, generative AI would need to drive a 15% increase in labor productivity just to justify current investment levels across the US economy. That number hasn’t materialized. Not in GDP data, not in corporate earnings calls, not in honest internal audits.
Meanwhile, the four biggest AI spenders, Microsoft, Google, Meta, and Amazon, collectively spent more than $300 billion on capital expenditures in 2025 alone, according to their public filings. The infrastructure is being built. The returns are being promised. The earnings reports still don’t reflect the claimed efficiency gains at scale.
Gartner projected that 30% of generative AI projects would be abandoned before reaching production by the end of 2025. That number tracks with what operators on the ground are actually reporting.
The Hard Truth About Who Wins and Who Doesn’t
Here’s the split most people won’t say out loud. There are two kinds of companies right now.
The first kind bought into the hype, hired a “Head of AI,” subscribed to a dozen tools, and told the board they were serious about AI transformation. They spent real money. They got real overhead. Their quarterly numbers look the same or worse because they never tied spending to a measurable output.
The second kind picked one specific, high-friction process, automated just that, measured the result in dollars or hours, and then decided whether to expand. These companies are actually seeing returns. Not because AI is magic. Because they treated it like a business decision instead of a branding exercise.
According to MIT research published in 2023, AI writing assistance improved individual task speed by roughly 40% for skilled workers in structured tasks. That’s real. But “individual task speed” is not the same as company revenue. The translation from tool efficiency to business outcome is where most organizations fail.
I’ve watched businesses spend six figures on AI subscriptions and see zero change in output per employee. The tools work. The implementation doesn’t. That’s a management problem, not a technology problem.
If you’re tracking where your AI budget is actually going, tools like Wallester make it easier to see spending patterns by category across your team. That kind of visibility matters more than the tools themselves when you’re trying to justify ROI.
What I Would Do If I Were Running Your Company Right Now
Stop thinking about AI as a budget line. Start thinking about it as a test. Every AI deployment should have one question attached to it: what measurable thing gets better in 90 days? If you can’t answer that before you buy, don’t buy.
The businesses winning with AI right now are not the ones with the most tools. They’re the ones with the clearest measurement systems. You need to know your cost per output before AI and after. If the number doesn’t move, the tool doesn’t stay.
Second, watch your headcount math. A lot of companies are adding AI talent while keeping the same legacy staff in the same roles. That’s how you end up with a larger payroll and the same throughput. If AI is genuinely handling a task, someone’s role should change. If no role changes, no real adoption happened.
For companies scaling a team around new AI workflows, making sure your payroll processes keep up with that change is practical. Gusto handles payroll for teams going through rapid role shifts without requiring a full HR department to manage it.
Third, ignore the headlines and read the line items. When a company says AI “improved productivity,” ask: in which department, by what percentage, measured against what baseline? Vague claims are a sign no one actually measured anything.
The Bottom Line
The Apollo economist is right. For most companies, AI is a cost center dressed up as a strategy. The ones who will look smart in five years are not the ones who spent the most. They’re the ones who measured first, spent second, and treated every AI deployment as a business test with a pass or fail outcome. The hype cycle is real. The ROI, for most, still isn’t.
Frequently Asked Questions
Why are most companies seeing no financial payoff from AI?
Most companies implement AI tools without tying them to a specific measurable output. The tools may work at the task level, but without clear baseline metrics and accountability, gains don’t translate to revenue or cost reduction. It’s an implementation and measurement problem, not a technology problem.
What does “zero financial payoff” actually mean for AI investments?
It means that when you look at company financials before and after AI adoption, the numbers don’t change in proportion to the spending. According to economists like Torsten Slok at Apollo, the macro productivity data simply doesn’t reflect the billions being spent on AI tools and infrastructure.
Are any companies actually making money from AI right now?
Yes. Companies that automate a single high-friction, high-volume process and measure the result in hard dollars are seeing real returns. The pattern is narrow deployment with clear measurement, not broad adoption across every department at once.
Should businesses stop investing in AI?
No. But they should stop investing without a measurement framework. Every AI spend should come with a 90-day test, a baseline metric, and a decision point. Spend with a hypothesis, not with a trend.
How do you measure AI ROI effectively?
Pick one process with a known cost per unit before AI. Implement one tool. Measure the same cost after 60 to 90 days. If the cost dropped and quality held, the investment works. If not, cut it and try somewhere else. Simple math, not strategy decks.
