Software agents that research, shortlist and purchase on a person’s behalf, which means the thing reading your product page is a parser rather than a shopper.
Also known as agentic shoppers, machine buyers, non-human customers, autonomous buyers
A machine customer is any non-human system that makes or materially shapes a purchase decision. The person still sets the goal (book a flight under four hundred dollars, restock the office coffee, find a moisturiser without fragrance) but the comparison, the shortlist and increasingly the checkout are handled by an agent working through APIs, structured feeds and product pages.
The distinction that matters is not that a robot is buying. It is that the buyer has no eyes. A machine customer does not register your art direction, your headline rhythm, or the fact that the hero shot cost forty thousand dollars. It registers fields: price, availability, dimensions, return window, certifications, review count. Anything you communicated visually and never encoded as data is invisible to it.
Forrester’s mid-2026 read on agentic commerce puts real money behind this. Projections have agentic shoppers driving up to 385 billion dollars in U.S. e-commerce by 2030. That is not a fringe channel, and it is not a channel that responds to persuasion in the way every existing agency skill assumes.
This is the first audience in the history of the industry that cannot be moved by craft, and agencies are structurally built to move audiences with craft.
Your product data becomes creative. If an agent cannot parse your feed, your brand is not in the consideration set, no matter how good the campaign was. The unglamorous work of clean attributes, complete specs and consistent naming is now upstream of everything else. Somebody at the agency has to own it, and historically nobody does.
Persuasion moves earlier in the funnel. You are no longer persuading the buyer at the point of purchase. You are persuading the human who writes the prompt, and then making sure the machine that executes it can find you. Those are two different briefs, and most agencies are only staffed for the first.
Measurement gets strange. Agent traffic does not behave like human traffic. No scroll depth, no dwell time, no creative fatigue curve, sometimes no referrer at all. If your reporting stack treats it as bot noise and filters it out, you will under-report a growing share of revenue and misprice the channel for your client.
A DTC skincare client sees checkout conversions climbing from a source the analytics team keeps flagging as suspicious: sessions with no scroll events, sub-second page times, and a straight line to cart. The account team wants to block it. Instead the strategist pulls the top twenty converting SKUs and finds the pattern: every one of them has a complete ingredient list in structured data, and every one of the non-converting SKUs describes its ingredients in a lifestyle paragraph and a photograph of a leaf. The fix is not a campaign. It is a two-week project to move ingredient, allergen and pack-size data out of the copy deck and into the product feed, which the agency has to invent a line item for because it does not fit media, creative, or production.
The automations and agents module of the workshop teaches you how to build AI workflows that compress the busywork without taking the craft out of the studio.