In the bullring, the moment of truth is the end of the afternoon: the matador stops performing for the crowd, squares up and goes in with the sword, and everything that came before is judged by what happens next. The Swedish service scholar Richard Normann borrowed the phrase for the instant "the service provider and the service customer confront one another in the arena," and his ideas helped Jan Carlzon turn around Scandinavian Airlines in the early 1980s. Carlzon made the phrase famous with a piece of arithmetic. Each of the airline's 10 million customers, he wrote in Moments of Truth, came into contact with about five employees a year, for about fifteen seconds each time. SAS, he concluded, was "created" 50 million times a year, fifteen seconds at a time.
The arithmetic was the argument. A brand built from 50 million small encounters is an average, and an average is forgiving. One surly gate agent is a rounding error in a sample that large; what matters is the habit of the whole organization, so the right investment is training and authority pushed out to the front line, where the encounters happen.
Marketing spent the decades that followed adding moments to that count. By 2005 Procter & Gamble had a director for what it called the first moment of truth: the three to seven seconds when a shopper notices an item on the shelf. Google added a zero moment of truth in 2011, the research a buyer does online before she ever reaches the shelf. Each new moment was another place to be present, and the brand became the sum of more and more of them.
The Count Is Falling
That count is now running in reverse. The research that once happened across a dozen tabs, a few vendor sites and a call with a salesperson increasingly happens inside a conversation with an assistant that reads the sources, summarizes them and hands the buyer a shortlist. The brand is still being evaluated. It is simply not in the room for most of the evaluation.
Gartner's newest survey of 408 senior marketing leaders, published Monday, shows marketers reacting on two fronts at once. Sixty-one percent say they have adapted their content so machines can read and represent it, and 42% say they have taken steps to monitor and correct misinformation about their brand in AI-generated answers. In the same survey, 94% plan to increase investment in human touchpoints over the next year. MediaPost read the same figures as a gap, and it is one: most marketing leaders are not yet checking what the assistants say about them. Gartner's analyst put the underlying shift plainly: as customers delegate more of the journey to AI, "the remaining human touchpoints become more consequential."
The buyer's side of the same shift is in Gartner's survey of 645 B2B buyers released in May. Two thirds of them, 67%, said they would prefer a buying experience without a sales rep, and 70% would prefer one that is entirely digital and self-service. Yet 69% said they prefer to validate what AI tells them with a sales rep. Buyers were also almost evenly split on whom they trust less: 51% said they were more likely to get misleading information from generative AI, and 49% from a sales rep.
Read together, those numbers describe a buyer who wants to talk to you less and needs you more when she does. She will skip the discovery call, the webinar and the nurture sequence. She will not skip the moment when she holds the machine's answer up to a person and asks whether it is true.
Small Samples Are Not Forgiving
This is where Carlzon's arithmetic turns against anyone who still runs on it. His brand was safe because it was an average of fifty million encounters. A brand that a buying committee meets in person four or five times across a deal has no such protection. In a small sample, every observation moves the mean. A pricing conversation that dodges a question, a demo run by someone who has not seen what the assistant told the buyer, an onboarding call that contradicts the summary the buyer brought with her: none of these is noise anymore. Each is a large share of the evidence.
The economics follow from the same fact. When human contact was plentiful, the sensible goal was coverage: more touches, more reps, more sequences, each one cheap and each one a small contribution. When contact is scarce, the marginal value of a single encounter rises, and so does the cost of getting one wrong. The budget question shifts from how many moments a team can afford to staff to how much each remaining moment is worth, and whether the person standing in it is equipped for what it now carries.
What it now carries is mostly verification. The buyer arrives with a summary written by a machine that has read your site, your reviews and your competitors' claims, and she wants a human to confirm it or correct it. A seller who does not know what that summary says cannot do either. This is why the work of reading and correcting what assistants say about the brand is not a separate visibility project. It is preparation for the conversation, as basic as knowing your own price list.
Staffing the Sword
It is tempting to read the shrinking count as permission to staff less: the assistant is doing the research, so the team can be smaller. That gets the arithmetic backwards. The assistant is taking the forgiving moments, the many small ones where a weak performance was diluted, and leaving the few decisive ones where it is not. The right response is fewer people in more consequential roles: the ones who handle pricing, technical validation, the first call after a purchase and the first problem, chosen and trained for those moments rather than spread across a funnel built for volume.
It also changes what leadership should measure. Touches per account and meetings booked made sense when the brand was an average. When it is a handful of observations, the measure that matters is what happens after each one: whether the buyer moved forward, whether the claim she came to check held up, whether the summary in her hand matched what she was told.
Carlzon's airline was created fifty million times a year. Yours may now be created a handful of times per buyer, and the machine will usually speak first. Make sure the person who speaks second knows what it said, because that is the moment the afternoon is judged.
