Who Owns the Gap Between What AI Tells Buyers and What You Can Deliver?

Boards tend to file AI search under marketing. Data from Infinity, a call analytics provider, suggests some of its effects land elsewhere. Its whitepaper, authored by Daniel Wilkinson, the company’s Chief Customer Officer, analyses 41 million phone calls from the past 12 months across a variety of UK sectors, and finds inbound volume still climbing: up 24% between the first half of 2025 and the first half of 2026. People are not abandoning the phone. What matters for a business is what happens when they use it.

Daniel said, “AI was supposed to reduce inbound call volume, but it actually hasn’t. What has changed is who calls, why they call, and what they expect when they get through.

This paper draws on data from 41 million inbound calls tracked across Infinity’s customer base. It shows what that shift looks like in practice, and what it means for every paid media and commercial team managing it.”

Intent is arriving. Sales are not.

Infinity’s Smart Outcomes tool processed 5.75 million inbound calls and verified 1.3 million of them as showing real intent to buy. Of that group, 349,000 (26.3%) reached a positive outcome. The paper treats this as a problem at the point of sale rather than in lead generation: the callers are real, and the conversations still fail.

Why they fail has changed. A year earlier, the leading barriers among sales leads that did not convert were decision-making at 20.8%, pricing uncertainty at 17.8% and product understanding at 17.7%. Twelve months on, in Q2 2026, mismatch over price and budget had moved to the top at 31.6%. Availability, which had barely registered in 2025, stood second at 18.3%, and a new category, product not offered, accounted for 12.8%. Product understanding had dropped to around 8%.

Infinity attributes the change to buyers researching options with ChatGPT or Perplexity ahead of the call, often picking up an indication of price along the way. When the figure quoted on the call differs, the call fails.

Two questions beyond the sales desk

Infinity’s own remedy for the closing problem sits with sales: it advises briefing call-handling agents for callers who arrive already holding a price. The findings also raise two questions that reach further, though these are inferences from the data rather than conclusions the paper draws.

Price. The paper explains that the quote on the call fails to match what an AI tool told the buyer. It does not establish where those tools get their prices or how accurate they are. A business that has not checked what ChatGPT or Perplexity say about its own offers cannot tell whether the gap lies in its quote, in the AI’s figure or in the buyer’s expectation, and finding out is a pricing and commercial task as much as a sales one.

Availability. Infinity attributes availability failures to the same mismatch with what AI-led buyers expect, but it does not say what causes the shortfall underneath, whether stock, capacity or lead times. For businesses managing stretched supply chains, that is a prompt to check whether what is published online about range, stock and delivery matches what operations can deliver. That is an inference, not something the paper tests

A budget question hiding in the attribution

A further question sits in how results are credited. Infinity says visits that click straight through from AI platforms can already be tracked, and argues that where they are not recorded as a channel of their own, the conversions that follow are given to whichever touchpoint came last, which it says is usually paid search or direct.

The channel is still small. Infinity traced 16,381 calls over the last 12 months to visitors who had clicked through directly from an AI platform, well under 1% of the 41 million calls in the study, though the rate of conversion events among them was 46% higher than for paid search referrals. Its referral-growth figures point to fast expansion, but they end in April 2025 and cannot say where the channel stands now. For a board, the question is less the size of any misattribution today than whether the business could see this channel at all if it kept growing.

What the data cannot tell a board

Four limits are worth stating plainly. Everything here comes from the customers of a single provider, and that base is UK-focused, so how far the figures carry across other European markets is untested. Infinity warns that its barrier categories were renamed when it moved to a new classification version, which means the broad trend is sturdier than any precise change in percentage points. 

The account of why buyers behave this way (that AI is pre-qualifying them before they call) is Infinity’s interpretation, and call records alone cannot confirm it. And none of the data speaks to margins or pricing power, so the commercial consequences for a business remain untested.

Wilkinson’s closing line is addressed to marketers: “If you’re a marketer still treating phone calls as an offline black box, you’re leaving your best first-party data on the table.” For a board, a practical first step is a modest one. Ask the commercial and operations leads to put the questions a buyer would ask ChatGPT or Perplexity, then compare the answers with what the business would actually quote and deliver. If the two diverge, the question in the headline needs an answer with a name attached.