The Confidence Gap: Why More AI hasn’t meant Faster Growth

Boards are certainly expecting AI to deliver growth – and that pressure sits hardest on enterprise and hyperscale companies, where budgets are the biggest and the scrutiny is closest. Gartner’s 2026 CMO Spend Survey has CMOs allocating over 15% of their marketing budgets to AI, with a clear message from leadership that it should act as a force multiplier for growth, efficiency and transformation. And the pressure to prove it is real – most marketing leaders say they’ve been asked to justify AI spend at board level, with only a small fraction of them feeling they can do that with hard evidence. 

When that ROI doesn’t materialise, the explanation usually offered is operational: fragmented systems, teams that aren’t ready, measurement that hasn’t caught up with the technology. All of which is probably true, and worth fixing. But it treats the gap as a plumbing problem: get the systems and training right, and the growth will follow. 

But there’s a simpler explanation sitting underneath that one, which none of these surveys are really set up to explore. You can give a buyer more information than they’ve ever had, but it doesn’t automatically make them feel safer about the decision they’re making. Those are two separate problems. Fixing the operational side gets you closer to the first but it does very little for the second – and AI, however good it gets, is only really built to solve the first. 

The evidence sits in how buyers actually behave 

Gartner published new findings a few months ago that are worth considering if you’ve put any real budget behind AI-driven growth this year. They surveyed 645 B2B buyers and found, well… roughly what you’d expect. 

Buyers are doing more independent research than previously, arriving more informed, and are often most of the way to a decision before a seller even knows they exist. Nearly half had used generative AI to gather information on vendors along the way. For years this has been the scenario marketing and sales leaders said they wanted: an informed buyer who needs less hand-holding to get to a ‘Yes’. 

Yet 69% of those same buyers still went and found a human to check the AI’s answer before they trusted it. 

Part of that is simple caution – AI still hallucinates often enough that buyers have learned not to take it at face value. But even an AI which never got a single fact wrong wouldn’t fully solve this, because being informed was never quite the same problem as being ready to stake your reputation on a decision. 

Why information was never quite the bottleneck 

There’s a reasonably well-established explanation for this, and it isn’t that buyers are being stubborn or that the AI gave a bad answer. It’s loss aversion – one of the more robust findings in behavioural science, and one that keeps turning up in new contexts whenever I go looking for it. We feel the pain of a bad decision more acutely than we feel the pleasure of a good one, and the higher the stakes, the more that can hold. 

A buyer signing off a six- or seven-figure contract isn’t just weighing features against price. They’re weighing what happens to their own reputation if it goes wrong – the credibility spent backing an unproven vendor, the conversation they’d have to have with their own boss if implementation stalled. An AI tool can give them a good answer but it can’t be accountable if that answer turns out to be wrong, and that’s really what the buyer is looking for when they go and find a person to talk to. It partly explains why the messenger tends to matter almost as much as the message – a confident-sounding AI response and a credible person saying much the same thing land quite differently, because only one of them can actually share the risk with you. 

It isn’t only a sales conversation 

Once you notice this pattern, you see it isn’t confined to the moment a rep and a buyer are talking. It shows up earlier in the buyer journey, where AI is increasingly shaping the shortlist, before anyone from the vendor side is contacted. It shows up in content and brand, where AI-generated material is making a lot of vendors sound much the same as each other, leaving human judgment as one of the few things still doing real differentiation. 

And that’s quietly changing what the sales motion itself is for – less about handing over information a buyer can now get anywhere, and more about being someone willing to put their own credibility behind being right. 

It shows up in three different places, but it’s one problem, not three – which means no single team can fix it by improving their own piece. Someone has to decide, deliberately, to spend the budget and the people on the human skills side of this, not just the technology. 

What this means for leadership 

None of this is an argument against investing in AI. The more useful question is whether the investment has quietly assumed that AI adoption looks after itself once the information gap closes. 

The evidence so far suggests it doesn’t, and that this isn’t a gap that closes just because the models get better. AI will likely keep getting better at gathering information. But it’s less clear whether it will ever be the thing a person trusts with their own reputation – and that gap, the human one and not the technology gap, looks like where the next real advantage in B2B sits.