The single most important number for any developer building AI into a product right now isn't a benchmark score or a token price. It's this: outside of the tech sector, AI has delivered zero net financial payoff to corporate America.
That claim comes from Torsten Sløk, chief economist at Apollo Global Management, who analyzed aggregate corporate data and concluded that the massive spending on AI infrastructure and software has not yet moved the needle on revenue or profit for non-tech firms. The finding was reported by TechCrunch.
This should change how you think about your next AI feature. The bottleneck isn't model capability. It's integration cost.
Why integration cost is the real constraint
Most developers focus on model accuracy, latency, and cost-per-token. Those matter, but Apollo's data suggests something deeper: companies are spending heavily on AI but seeing no return because the surrounding system — data pipelines, workflow redesign, user training, compliance — eats the entire budget before any value materializes.
If you're building an AI product for a non-tech enterprise, your biggest risk isn't that the model isn't smart enough. It's that the total cost of making the model actually useful in a specific business process exceeds the value it generates.
What developers can do differently
Measure deployment cost, not just inference cost. Track how many engineering hours are needed to connect the model to existing databases, APIs, and approval workflows. That's your real cost.
Ship a narrow, high-certainty use case first. Apollo's analysis implies that broad "AI assistant" pitches fail because the integration surface is too large. Pick one process where the output is easy to verify and the user already has a clear pain point.
Assume the model will be a commodity. The economist's finding suggests that model quality alone won't drive business value. The advantage comes from how cheaply and reliably you can embed the output into an existing decision loop.
The lesson from Apollo's data is uncomfortable but useful: if your product's value proposition depends on the AI being "better" rather than on the integration being cheaper than the alternative, you're likely building for a market that doesn't yet exist.
Source: TechCrunch - AI delivers 'zero financial payoff' to most companies, Apollo economist says
Top comments (0)