AI didn't kill your sales motion. It exposed that you never had one.

Why do AI sales tools fail?

The best funded AI SDR company in its category lost most of the customers it won. ZoomInfo, the buyer, ran it against its own reps for a month. The reps it was meant to replace closed more deals than it did.

I work with B2B founders at the 5 to 100 person stage who run their sales motion themselves, and I've watched more than one of them live a smaller version of this story inside their own AI stack. Not a public collapse. A quiet one. A tool that promised to do the selling and instead did the typing.

AI did not kill anyone's sales motion. It ran straight through the hole where the motion should have been, and the hole is now visible to the board too. The easy read, that a story like the SDR failure proves AI cannot sell, misses what actually happened.

Why do enterprise AI projects fail?

The pattern is bigger than one company. RAND interviewed 65 data scientists and engineers about why AI projects fail. Eighty four percent named leadership decisions and expectations as the primary cause, ahead of anything technical. That is an opinion survey of practitioners, not a measured failure rate. It still points the same way. The model worked, and there was nothing underneath it to run.

Source: RAND, The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed, August 2024.

Daniel Käfer's read on the sales function draws the same line more sharply. The roles getting compressed are the mechanical ones: outbound sequencing, lead routing. The roles surviving are judgment calls: qualifying a real opportunity, negotiating a hard deal, deciding which conversation earns a founder's hour this week. AI is good at the first list. It cannot do the second one, because the second list was never a task to automate. It was a discipline someone had to build first.

What should you install before your next AI purchase?

Pull the last five deals your AI tooling flagged as hot this month. For each one, ask what the buyer loses by doing nothing this quarter. If your tool cannot answer that, and neither can you, the tool was never wrong. It was reporting on a motion that was not there to measure.

The founders whose AI tools are actually paying off started before the AI, writing down what a real opportunity looks like, what a stalled deal looks like, and what should happen at each stage of a serious conversation. Once that existed, the AI had something to amplify. Before that, it had nothing to point at.

That written down version of the motion is the revenue operating system underneath everything the tooling promises. Install it first and the AI looks smart. Skip it and the AI just runs the failure faster, with better reporting attached.

Käfer is reading the org chart. Ray Arain is reading the RAND and Gartner numbers. Both landed on the same finding in the same week: install the motion first, or watch the tooling run straight through the hole where it should have been.