Opinion · Channels · Audit by construction in agent commerceAfter search
Marketing bought clicks for twenty-five years. The agent doesn't click.
The regulator no longer just reads the ad. It reads every decision.
A customer is looking for travel insurance for a family trip. Two years ago, they would have typed the query into a search box, scanned three sponsored results, clicked the one with the best framing, filled the form, bought. Today, the same customer asks their personal agent. The agent is not searching. It is running a regulated negotiation across insurers: real-time offer, eligibility decision, reason for the offer, binding commitment, audit log. The whole transaction takes ninety seconds. The insurer that can decide audit-grade in real time wins. The insurer that publishes a static brochure becomes a regulatory liability the agent will route around. ChatGPT, Claude, Gemini and Perplexity together cross half a billion weekly active users. The channel has been carrying real traffic for two years; the regulator is catching up.
Why agent commerce is a regulated decisioning channel
Search advertising was regulated for the claim. The brand was responsible for what the ad said: truthful, substantiated, disclosed. Agent commerce extends regulation to the decision. The brand is now responsible for what the agent decided on the customer's behalf, on what basis, with what audit trail.
Five regulatory anchors apply or will apply within an eighteen-month window: the EU AI Act high-risk provisions on agent decisioning, the RBI FREE-AI Framework, the CBUAE Guidance Note on AI-mediated customer decisions, SAMA's annual revision, and FTC enforcement on AI-mediated marketing claims. The brand's interaction with the customer's agent is no longer a marketing event. It is a regulated decision.
Operators that built audit-grade decisioning for an earlier regulatory cycle already know the shape. Banks under SR 11-7. Insurers under Solvency II. Regulated service providers under conduct rules. They built the audit trail as a by-product of the decision, not an export from it. They are the operators whose agent commerce participates from day one.
What the agent and the regulator both ask
The agent's queries and the regulator's requirements are the same questions. Both want to know the brand's offer for this customer today, on what basis, with what eligibility, with what audit-grade reason code. Both want to know who is accountable. Both want the answer in real time, and logged.
Most consumer brands cannot answer this way today. They published a static API for the human-search era and have not invested in real-time decisioning or audit-grade reason codes since. The agent treats their offers as un-decidable. The regulator will treat their unaudited decisions as non-compliant.
| The question | Static-publishing brand answers | Audit-grade decisioning brand answers |
|---|---|---|
| What is your offer for this customer today? | The published price. | A real-time decision factoring customer cohort, eligibility and constraints. |
| Why this price, on what basis? | The price is published. | Reason code, model, inputs, eligibility logic. Exportable on demand. |
| Will you commit? | The website terms. | A binding commitment with the audit log to prove it. |
| Who is accountable? | The vendor's T&Cs. | The operator, named in the decision log. |
| Can the customer or regulator audit? | No machine-readable trail. | Full reason-coded decision log in the regulator's format. |
Search regulated the claim. Agent commerce regulates the decision.
The brands the agent queries
The agent's consideration set is the brands that can decide and prove the decision. The brand that publishes a price gets read once. The brand that decides in real time, with an audit-grade reason code, stays in the set across queries. The brand whose decisioning is a vendor black box is excluded by the regulator before the agent excludes it. The brand that has been buying its way into the human's consideration set with paid search is, in the agent's reading, not in the set.
What the channel becomes
The search channel, for the brand, is no longer a placement problem. It is a regulated decisioning problem. The brand's agent tech is its real-time decisioning capability and its regulatory posture at the same time. Three properties of agent tech that participates.
Real-time decision authority. The brand decides, in the moment of the agent's query, what offer to make on what eligibility with what commitment. Not a price published last quarter.
Reason-coded outputs. Every decision is logged with the reasons, in a format the customer's agent can read and the regulator can audit. Not a contractual disclaimer.
Operator-controlled change logic. When regulations, products or pricing change, the operator changes the decisioning. Not the vendor's roadmap.
Audit by construction is what operators with a regulatory track record have always called the architecture. The agent era is what makes everyone else have to learn the term.
What follows
The agent tech that participates in regulated agent commerce is unlike most current MarTech. Decisioning and execution happen in the same layer, so the agent reads the brand's offer and the basis for it together, in the moment of the query. Reasoning is produced as a by-product of every offer, in the regulator's format, so the audit trail is what the agent reads and what the regulator audits. The brand's customer record and decision trail stay with the operator, not the vendor. The pricing aligns to outcomes the agent verifies, not seats licensed. And the tech was built for regulated enterprise from the start, which is why the audit-grade discipline is already there.
The platforms that carry the brands through this shift were architected for the audit from day one, not retrofitted to a regulation. The regulation is catching up to the architecture, not the other way round. The brand whose agent tech has this character stays in the consideration set. The brand whose does not is un-decidable, un-auditable, and invisible.