Gartner's strategic predictions for 2026 and beyond include the expectation that by 2028, 90 percent of business buying will be intermediated by AI agents, pushing more than 15 trillion dollars of spend through agent exchanges.
Numbers of this kind do real work in boardrooms. They also belong to a class of evidence that deserves its own handling, and the handling is not to dismiss them.
Three questions we put to a forecast
What would have to be true. A figure like 90 percent requires standards, integrations and a settled question of who is liable when an agent commits a purchase. Those are visible, datable things. Their absence is informative and does not require an opinion about the technology.
What counts as the event. Intermediated is doing a great deal of work in that sentence. It can mean an agent recommending a supplier, or an agent completing a transaction. Those are years apart in difficulty and the forecast is far more plausible under the first reading.
What decision would change. If your answer is the same whether the figure turns out to be 90 percent or 30 percent, the forecast is interesting rather than useful, and it should not be in the paper you are writing.
Why we still take it seriously
The direction is corroborated by measurements rather than forecasts. Buyers already prefer a rep-free research phase. Answer engines already mediate a growing share of what a buyer sees. Sellers are already starting their research inside a model.
A forecast that points the same way as three independent measurements is worth acting on directionally, and the actions it implies are mostly things worth doing anyway. Making your commercial facts machine-readable, keeping scope and price stated in plain terms, and being present where a summary is assembled all pay off whether the figure lands at 90 or 30.
What we take from it
We use the weaker word deliberately on our own claims, and we apply the same test to other people's. This one is a hypothesis with a date on it, published by a firm with a strong record and a commercial interest in the subject. All three of those are true at once.
The response we would defend is preparing for the direction while refusing to budget against the magnitude. That is a less satisfying answer than either enthusiasm or dismissal, and it is the one the evidence supports.