Agentic Commerce: When Your Next Customer Is an AI

A shift is underway in how buying decisions get made, and it is easier to prepare for than to catch up with. Increasingly, the first thing to evaluate your business is not a person. It is software acting for one.

What agentic commerce means

A buyer describes what they want to an assistant. The assistant researches options, compares them against the stated constraints, and returns a short list — or, increasingly, completes the purchase. The human sets the criteria and approves the outcome. The middle is machine work.

For a supplier, that middle is where you are either included or silently excluded, and you get no signal either way.

What agents optimise for

Not brand warmth. Not a well-written value proposition. Agents work with what they can parse and verify: price, specification, availability, delivery time, terms, and third-party corroboration.

This inverts a long-standing assumption. Vagueness has historically been a sales tactic — get them on a call before mentioning price. Against an agent, vagueness is disqualifying, because a missing field is not a mystery to be resolved later. It is a row that cannot be compared.

What to do about it

Publish prices in a machine-readable form

Real numbers on the page, in Product or Service schema, and ideally in a plain-text file an agent can read without rendering anything. From $250 is a comparable fact. Contact us for pricing is an empty cell.

Make specifications unambiguous

What is included, what is not, timelines, service areas, terms. Anything a human would ask on a call, an agent needs written down. The exclusions matter as much as the inclusions — an agent that cannot verify a constraint will assume you fail it.

Keep availability current

Stale stock or booking data is worse than none. An agent that recommends you and then finds the thing unavailable produces a poor outcome for the buyer, and that is the kind of failure systems learn from.

Earn third-party corroboration

Agents weight independent sources heavily because your own site is a claim and a review platform is evidence. Directory listings, reviews, and mentions in reputable roundups do more for machine confidence than another page of your own copy.

What has not changed

Agents shortlist. Humans still choose, and they still choose on trust, proof of work and whether they believe you will do a good job. The shortlist is a filter, not a decision. The change is that you now have to pass a machine filter before a person ever sees you — and passing it is a data exercise, not a copywriting one.

Where to start this quarter

1. 1Publish a real starting price for every service. One afternoon.
2. 2Add Product or Service schema carrying those prices.
3. 3Write the exclusions and constraints down explicitly.
4. 4Check an assistant can already describe what you sell and what it costs. If it cannot, that is your gap.