A Gartner survey finds that most B2B buyers now check AI-generated insights with a salesperson. The first analysis is done before the seller arrives. What remains is the part only a person can do: stand behind it.
When we launched the working paper we led with one claim: AI is not making human selling obsolete, it is making generic human selling less valuable. A survey published by Gartner in May gives that claim a number. Sixty-nine percent of B2B buyers prefer to validate AI-generated insights with a sales representative. The buyer has already asked the machine. They come to the seller with an answer, and they want to know whether it holds.
What the buyer is asking for
Read the finding from the buyer’s side. They have a draft analysis, a shortlist, perhaps a first business case, all produced in minutes. What they do not have is anyone accountable for it. An AI tool will not co-sign an assumption in front of the CFO, will not notice which of four stakeholders is still hesitating, and will not be in the room when something goes wrong after signature. So the buyer brings the draft to a person and asks, in effect: would you put your name to this? That is a request for judgement and accountability, not for information. It is the same request our interviewees describe when they say trust has moved from before the sale to after it.
What it changes for suppliers
Two things follow. First, the seller who arrives with a presentation is arriving with what the buyer already has. The value is in disagreeing well: showing where the AI’s baseline is wrong, which assumption the buyer’s own finance team will not accept, and what the analysis missed about their system. Second, validation is only worth something if it comes with consequences. A seller who validates a conclusion and then walks away has added little; one who validates it and stays accountable for the result has added what the machine cannot. That is why the research keeps returning to the fourth and fifth demands: show the evidence in the customer’s numbers, and guarantee the results. Validation without a guarantee is a second opinion; with one, it is a proposition.
The honest qualifier
A survey measures a preference, not a behaviour: buyers say they want a person to validate the machine, which is not the same as paying for that validation. Gartner’s respondents are also weighted towards large organisations and technology purchases, where AI-assisted research is furthest advanced; the pattern may be weaker in industrial and service markets for now. And a supplier of sales research should be careful about welcoming evidence that sales still matters. What the number supports is narrower than the headline: buyers do not want less human contact, they want a different kind. Whether commercial organisations can deliver that kind — diagnosis, judgement, accountability — at scale is exactly what the paper says is not yet proven.
The work
If seven in ten buyers arrive with an AI-generated answer, the first question in every sales conversation has changed. It is no longer “what do you need?” It is “what have you concluded, and what would it take for me to stand behind it?” Commercial teams built to present will struggle with that question. Teams built to engineer and prove value will recognise it as the job.
Is your proposition built to be validated, or only presented? Take the seven-question self-audit. From the Future of Selling research programme, co-led by Prof. dr. Régis Lemmens and Prof. dr. Javier Marcos — see the working paper Engineering Systemic Value in B2B.