Why won’t buyers give honest feedback to an AI?
Buyers won’t give honest feedback to an AI because the bottleneck in win/loss has never been scale, it’s the honesty of the conversation. Sending an AI to debrief a buyer after a significant purchase decision signals that scale matters more than the relationship, and it skews responses toward buyers with the least complicated stories to tell.
The relationship signal matters
There’s a version of AI-assisted win/loss that sounds efficient on paper: automate the outreach, collect feedback at scale, run analysis on the responses, no scheduling overhead. The problem is that buyers in B2B decisions weigh human interaction as much as the technology itself. A buyer who just made a six-figure purchase decision and receives an AI-run debrief request experiences something close to the frustration of a customer service bot standing between them and a real answer: they know the information exists, they know a person could hear it, and they’re being routed into an automated loop instead.
That experience tells the buyer something about how the vendor values the relationship, and it shapes how much the buyer is willing to share in return.
It skews the data before a single answer is collected
The effect shapes who responds at all, not just how they respond once they do. Buyers willing to engage with an AI-run survey or interview are not a representative sample of the full buyer population. They tend to be the ones with low friction, low stakes, or low candor about the decision. Buyers who had the most uncomfortable, complicated, or internally contentious reasons for their decision are the ones most likely to decline the interaction entirely and move on, taking the most valuable feedback with them.
AI still has a role, just not this one
AI has real leverage in win/loss research, downstream of independently gathered interviews, where it’s genuinely useful for pattern recognition and synthesis across an honest dataset. But the honest conversation has to happen first, and that conversation requires a buyer to sense that a real person is listening, has no stake in what they say, and is genuinely trying to understand what happened. A tool that automates the collection step makes the problem worse, by removing the one element, human trust, that gets buyers to say the thing they’d never say to the vendor directly.
What this means for a win/loss program
Buyers will tell someone the truth, if that someone has the empathy to listen, the context to probe, and the ability to make the buyer feel safe enough to say what they’d otherwise leave out. An independent human researcher can be that someone. An AI interviewer, no matter how well designed, cannot yet replicate the trust that makes candor possible.
This isn’t primarily a technology limitation that a future model release will fix. Buyers withhold candor from AI interviewers for the same reason they withhold it from the vendor’s own sales team: because the entity on the other end of the conversation has a stake in the relationship, or is perceived as an extension of one. A vendor-deployed AI interviewer inherits that perception the moment a buyer recognizes what it is. Until an AI system can convincingly present itself as having no connection to the vendor and no stake in the outcome, which raises its own set of trust and disclosure problems, the honesty gap it faces will track closely with the honesty gap a vendor’s own team already faces when asking buyers directly.
The practical implication for a GTM leader evaluating research options is straightforward: candor requires perceived neutrality, and perceived neutrality currently requires a human, working independently of the vendor, to be the one asking the questions.