For years the rule in the software business was simple: you sign a three-year contract, the client gets locked into your system and the revenue flows on its own. That model held up the entire sector. A new set of research shows that artificial intelligence has broken precisely that model - and much faster than anyone wanted to admit.

The numbers contradict each other. Of 150 IT professionals surveyed at large companies, 74 percent plan to increase their AI budgets over the next twelve months. At the same time, fewer than half of all pilot projects ever reach full deployment. The money is growing, the results are not.

The key figure is the third: 77 percent of companies reassess their AI vendors every six months or continuously. Research by the Madrona fund calls this a "fast in, fast out" dynamic and says it is fundamentally different from the classic software business, where multi-year contracts functioned as a moat around revenue. That moat is now gone. Or, as they write: the cost of switching vendors is lower, and the pace of reassessment is merciless.

What does that mean in practice? A startup boasting fifty million in revenue today cannot guarantee those same clients will be there in six months. And valuations are built on precisely the assumption that they will. When an investor multiplies annual revenue by twenty or thirty, what they are really buying is an assumption of durability - and that assumption is now the weakest link.

The context makes this more interesting still. According to IDC projections, companies will spend 4.25 trillion dollars on technology in 2026, largely because of artificial intelligence. But an MIT report from last year found that 95 percent of corporate AI projects returned nothing on what was invested. An enormous river of money flowing through a channel with a hole in the bottom.

Which is why the billing model is changing too. Of 50 technical buyers surveyed, more than half want to pay by outcome - by what the tool actually completed - rather than by tokens consumed or number of users. Andreessen Horowitz partners Tuğçe Erten and Sarah Wang call this billing that is economically valuable to both sides. Translation: clients no longer want to pay for a promise.

Is this bad news? It depends which side of the table you are sitting on. For a firm that genuinely delivers - better, because the result now gets paid for directly. For a firm that lives off presentations - very bad, because six months come around fast.