Seventy-six percent

Every conference stage this year has an agent on it.
Here is a figure that was not on any of those slides: roughly three quarters of purchases in 2026 still happen with a person standing in a shop. Seventy-six percent, by one count. (Mirakl)
I am not telling you that to be comforting, and I am not telling you the agents do not matter, I wrote a whole piece saying they do. I am telling you because the distance between where the industry’s attention is and where the industry’s money is has a name in our trade.
It is called an opportunity. And it closes.
Two true things that never share a paragraph
Put these next to each other, which almost nobody does:
More than 70% of shoppers already use a large language model somewhere in their shopping journey. (Mirakl)
And 76% of purchases still happen in person. (Mirakl)
Both figures are real. They are not in tension, and treating them as a contradiction is the first mistake. She is researching on a screen and buying in an aisle. That is not a transitional state on the way to something else. For a very large number of categories it is the settled behaviour, because a screen is good at comparing and a shop is good at having the thing now.
The mistake I keep watching teams make is reading the first number as a prediction and the second as a legacy. They are both descriptions of the present, and the present is where your quarter happens.
What I have seen this cost
I will give you the version with the numbers stripped out to protect the client, because the shape is what matters and the shape repeats.
A team reallocates hard toward a new channel on the strength of its growth rate. The growth rate is real, it is 200% of something small. Meanwhile the thing that was funding the business gets described internally as mature, which is a word that means we have stopped thinking about it. Investment drifts. Execution in store slips, quietly, in the way execution always slips: a display goes up late, a planogram compliance check stops being run, a shopper-facing message gets approved that nobody tested with a shopper.
Eighteen months later the new channel is genuinely working and the base has leaked more volume than the new channel ever added.
Nobody made a stupid decision. Every individual step was defensible. The error was structural: they let the growth rate of a small thing outrank the absolute size of a large one, and no single meeting was where that happened.
This is the oldest failure in commercial life and the newest technology does not change it.
The tools change. The shopper does not.
Here is the line I keep coming back to, and it is not a comfort, it is a discipline.
The machines are genuinely remarkable now. They read shelves from photographs. They tune prices faster than a human can approve them. They will, increasingly, place the order themselves. All real.
None of it changes how a human being decides.
She is still tired. She is still hurrying. She is still watching a budget, and she still gives your brand something in the region of nine seconds before her hand moves. The three gates still hold: she has to see it, it has to feel safe, and the buying has to be easy. Those are facts about a nervous system that has not been updated since long before any of this, and they will not be updated by a protocol.
Technology moves fast. Psychology barely moves at all. That gap is where every durable advantage in this discipline has ever come from.
What to do about it
Fund the present properly. The 76% is not a legacy business, it is the business. If your in-store execution has quietly degraded while everyone was in agent workshops, that is not a strategy problem, it is a housekeeping problem, and it is the cheapest volume you will find this year.
Judge new channels on absolute contribution, not growth rate. Ask for the number in units and currency alongside the percentage. A team that cannot produce it is selling you a slope.
Watch the in-store medium get interesting. This is the part I would actually spend new money on. In-store media is moving from static signage to measurable networks: digital screens, smart shelves and audio that update in real time on inventory, weather and time of day. (Mirakl, Rockbot) That is the aisle acquiring the capabilities the internet has had for twenty years, aimed at the 76%. It is the least fashionable thing in this article and I suspect the most valuable.
Stop asking which era we are in. You are in both. Everyone selling you a single answer is selling.
The nine seconds are not over
The agent era is coming and I have written about why it matters and what it breaks. But it is arriving into a world where most of the deciding still happens in front of a shelf, by a person with a basket, in the small window between noticing and reaching.
Everything my trade knows is about that window. It has not closed. Most of the industry has simply stopped looking at it, which, if you are the one still looking, is the best news you will get this year.
Sources
- Top Retail Media Trends for 2026: AI & Agentic Commerce. Mirakl. The 76% in-person figure, the 70%+ LLM usage figure, and the in-store media shift.
- Retail Media Trends 2026. Rockbot
- Retail Media Trends and Outlook for 2026. Tinuiti
Shopper Marketing in the Era of AI is built on this idea: the tools change, the shopper does not.