Most business coverage of artificial intelligence still starts in the wrong place. A typical story opens with what a new model can do-its benchmark scores, its context window, its multimodal capabilities-and only later, if at all, gets to the question that actually determines whether a business should care: does the economics work.
The gap between capability and value
A model's technical capability and its business value are related but not the same thing. A system that can pass a bar exam or write competent code is impressive; whether it can do so more cheaply and reliably than the process it's meant to replace, at the volume a business actually needs, is a separate and much harder question. Enterprises that have run serious AI pilots consistently report that the gap between a compelling demo and a production-ready workflow is where most of the real cost and complexity lives-integration, data quality, exception handling, and the ongoing cost of monitoring outputs for drift and error.
What business coverage should actually track
Useful AI business journalism asks a narrower set of questions than technology journalism does. What is the fully loaded cost of a deployment, including the human oversight it still requires? What specific role or process is it replacing, and what did that cost before? What happens when the model is wrong, and who is accountable for catching it? These questions don't require access to a company's internals-they can often be answered by looking at procurement patterns, job postings, and the pace at which pilots convert to renewed contracts.
Why the P&L lens matters more now
Early in any technology cycle, capability is the story because capability is scarce. As foundation models commoditize and the gap between leading vendors narrows, the business story shifts from "what can this do" to "who is capturing the value." That's a question about pricing power, switching costs, and where in the stack margin actually accumulates-infrastructure, model providers, or the businesses applying AI to a specific workflow. It's a more useful lens for investors, operators, and competitors alike than another round of capability comparisons.
A more disciplined beat
None of this means capability doesn't matter-it's the raw material every business case is built on. But treating capability as the whole story, rather than an input to a cost-benefit calculation, is how AI business coverage ends up amplifying vendor narratives instead of testing them. The businesses actually succeeding with AI right now are the ones asking the P&L question first.