In 2011, Google published a report that proved consequential for marketing. It was called ZMOT: Zero Moment of Truth. The central observation was disarmingly simple: purchase decisions do not happen at the shelf. They happen the moment someone begins to search. Whoever does not exist at that moment, whoever is not found, compared, or recommended, has already lost the race before it began.
That was a watershed in 2011. Marketing had to understand that the moment before the purchase has its own space, with its own rules, its own mechanics, its own winners and losers. Search engine optimization, content strategy, and review management suddenly stopped being technical side-shows and became the main stage.
What nobody anticipated in 2011: that was not the last shift of this kind. It was the second to last.
The Moment Before
We are currently moving into a phase where a growing share of purchase decisions are no longer made by people but delegated to AI agents. Not in any science-fiction sense, but in a very practical one: people give their agent a task, define their parameters, and the agent executes. Which hotel. Which software. Which supplier.
The decisive moment in this architecture is no longer the moment before the search. It is the moment before the prompt.
What a person passes to their agent as a fixed preference, a set constraint, an unquestioned assumption, that is the decision. Everything after that is execution. The agent does not search for the best option in an open space. It searches for the best option within the parameters it was handed. And what does not exist as a parameter does not exist for the agent as a preferred option.
This is the Zero Moment of Agency: the moment in which a person decides with which assumptions to send their agent out. That moment is the new strategic location.
A brand that is not anchored as a fixed parameter in a person's mind at the time of the ZMOA exists for the agent only as a candidate among many, evaluated by objective criteria. In that competition, it is not the strongest brand that wins. It is the cheapest offer.
What an AI Agent Does Not Have
At this point it is worth being very concrete, because otherwise the implication sounds too abstract. The question is: on what do classic loyalty programs actually run? What is their mechanism of effect?
The honest answer is: they run on human cognitive biases. Loss aversion, the psychological pain of losing something one already has, the half-stamped card, the points about to expire. Status-seeking, the desire to be visibly placed in a hierarchy, the gold tier, the platinum card, the exclusive lounge. Habit inertia, the tendency to continue familiar patterns because switching costs cognitive effort.
These mechanics work because human decisions are not purely rational. They are shaped by emotional, social, and evolutionary patterns the brain has developed over millions of years. Building a racetrack precisely calibrated for those patterns is the business model of classic loyalty design.
An AI agent has no loss aversion. It has no status needs. It has no habit inertia. It does not ask whether the loyalty card is already half full. It asks why it should be with this brand at all.
Building a racetrack designed for human irrationality, then being surprised when an algorithm is faster, is not a failure of execution. It is a structural problem: the algorithm simply does not perceive the obstacles the racetrack was built around.
What Survives the Filter
The relevant question then becomes: what survives an agent's filtering? What is agent-resilient?
The answer is: only what already exists as a preference in a person's mind before the agent becomes active. Not as passive brand awareness, but as a set parameter. As an assumption that is no longer questioned. As part of the brief, not the search result.
There are three concrete forms through which a brand can achieve that status. They are not program features. They are preconditions.
The first is genuine behavioral integration. The brand is so deeply embedded in daily routine that its removal would create real behavioral effort, not just a feeling of loss. The person does not switch away because switching is work, and because that work bears no reasonable proportion to any gain. This is not lock-in through contract. It is lock-in through actual utility that proves itself daily.
The second is identity anchoring. The use of the brand is part of the person's self-image. They do not ask their agent which motorcycle is objectively better. They ask which motorcycle they are. That question is agent-resistant because it accepts no answer based on data comparison. No algorithm can retroactively ensure that someone associates a brand with something meaningful. That work has to be done beforehand, in real experience.
The third is cumulative contextual knowledge. The brand has accumulated enough person-specific knowledge over time that switching produces a genuine quality loss. Not inconvenience, but actual loss of something that works and would need to be rebuilt elsewhere. The agent will find it harder to recommend switching when it sees that the existing data foundation has a value no competitor can replicate in the short term.
The Race to the Bottom
This is where the argument becomes uncomfortable. Not because the concept is new or difficult to grasp. But because the implication leads directly into the room where most marketing decisions are actually made.
Agent-resilient anchoring is expensive. It is slow. It does not show well in a quarterly report. Discounts and points are cheap, fast, and produce immediately measurable uplift. As long as marketing teams report quarterly, as long as campaigns are judged by conversion rate, as long as the next forecast dominates action, companies will prefer the measurable over the effective.
The result is a Race to the Bottom. The industry optimizes at full speed for a mechanic that an algorithm handles better than any campaign. And the further AI agents penetrate purchase decisions, the faster the ROI of that mechanic collapses. Not due to poor campaign quality. Due to the structural shift in the decision space.
This is the real resistance to ZMOA-oriented design: not missing knowledge, but misaligned incentive structures in the organizations that would need to implement it. The problem is not a knowledge problem. It is an incentive problem.
The One Uncomfortable Question
At the end of this argument sits not a solution but a question. A question every marketing team should answer for itself before planning the next points program, introducing the next status tier, or optimizing the next push notification.
The question is: if we turned off all discounts and points tomorrow, what would our customers actually miss?
Not what they would say. What they would actually miss in their behavior. Where habit would pull them. What would be absent from daily life because it was genuinely useful. What is connected to their self-image, not to the price.
Whoever can give a precise answer to that question has the foundation for ZMOA-resilient brand anchoring. Whoever hesitates knows what they are actually dealing with: a program built on human cognitive biases, and the question of whether an algorithm will soon exploit those biases more efficiently than their own marketing department.
The honest answer to this question is uncomfortable. That is intentional.