The headline numbers on AI have turned grim, and they agree with each other. MIT’s NANDA study found that ninety-five percent of generative AI pilots produced no measurable impact on the P&L, against thirty to forty billion dollars of spend. Gartner expects sixty percent of AI projects to be abandoned through 2026, mostly because the data was not ready. RAND puts the historical failure rate above eighty percent, roughly twice the rate of ordinary IT projects. Different researchers, same verdict.

The failure is decided before the project starts

Read past the headlines and the cause is consistent. It is almost never the model. It is the data the model is asked to run on, and the workflow it is dropped into. Gartner’s own definition of AI-ready data is data that is governed, pipelined, and tied to a specific use case. Most businesses do not have that. They have five tools that do not talk and a process that lives in people’s heads. You can put the best model in the world on top of that and it will still fail, because the failure was baked in before the project began.

The money is going to the wrong place

The MIT work has a second finding that gets less airtime and matters more. Most AI budget goes into sales and marketing pilots, and that is exactly where the return is lowest. The highest return sits in the back office, in the unglamorous workflow nobody wants to map. The projects that work are deeply integrated, built for one vertical, and usually delivered by a specialist who knows the work. Vendor-led builds succeed about twice as often as internal ones.

The model was never the problem. The business underneath it was not ready to absorb one.

What ready actually looks like

Ready is not a bigger model or a better prompt. Ready is the work running on a system instead of in someone’s memory. It is the workflow captured, the exceptions encoded, the data flowing into one place that can act on it. That is the part N+1 does first, before any AI goes near the business. Do the work, codify the workflow, capture the data. Then, and only then, does the model have something worth running on.

The opportunity in the failure

The eighty and ninety-five percent numbers are not a reason to wait. They are the opportunity. While most of the market burns budget on pilots that were never going to land, the operators who fix the data and the workflow first will be the few that show real P&L impact. The gap between the two is the whole game.