Problem
E‑commerce is designed for humans: search, compare, read, filter, add to cart and pay. But if an agent starts to investigate, compare and execute delegated purchases, many current pieces fall short.
An agent needs to understand constraints, availability, returns, permissions, budget, identity, payment method and confirmation. It doesn’t buy because a button looks nice. It buys because the flow is verifiable.
Thesis
Agentic Commerce will not be a cosmetic chat layer. It will be a restructuring of product, data, payments and trust for delegated decisions.
A store that wants to sell to agents will need readable catalogs, machine‑readable policies, purchase mandates, proofs, limits and payments that distinguish human intent from agentic execution.
Framework
An agentic commerce flow has six contracts:
- Mandate: what the human authorizes.
- Product truth: verifiable product data.
- Policy: returns, warranty, privacy and restrictions.
- Budget: economic limit and conditions.
- Payment: credential and authorization.
- Receipt: proof of decision, purchase and delivery.
Mini‑case: an agent purchases software for a team. It should not “click” like an invisible user. It must operate with a mandate, budget, criteria, approval when applicable and a traceable receipt.
Measurable signal: percentage of the catalog and policies that an agent can understand without interpreting images or ambiguous copy.
Position: if the buyer is an agent, your checkout becomes a trusted API.
Why It Matters Now
Google has shown more agentic experiences in Search for investigating and assisting tasks. In payments, the Agent Payments Protocol proposes an open way to connect agents, users, merchants and payment providers with verifiable mandates. Stripe has also published work around payments for agents and automated commerce.
The combination is powerful: agents that discover, compare and purchase with permissions.
That forces growth, product, legal and finance to speak the same language.
Anti-example
“We put a chatbot that recommends products.”
That is conversational assistance, not agentic commerce. The real leap occurs when the agent can act with a mandate, pay with control and leave evidence.
Protocol (3 steps)
- Make the critical machine‑readable. Product, price, stock, restrictions, returns and warranty.
- Design mandates. What an agent can purchase, up to how much and with what approval.
- Record evidence. Criteria, alternatives, authorization, payment and receipt.
| Contract | Question | Area |
|---|---|---|
| mandate | who authorized | legal |
| budget | how much can spend | finance |
| policy | what conditions apply | product |
| payment | how it is executed | payments |
| receipt | what evidence remains | operations |
Related
- AEO Ops: from SEO to operations for response engines
- Agentic Sales Signals: how agents evaluate your offer
- The Invisible Funnel: what disappears from the funnel when agents qualify before sales
Sources consulted
- Google: Search at I/O 2026
- Google Cloud: Announcing the Agent Payments Protocol
- Stripe: Agentic commerce guide
Next step
Review your purchase flow as if the customer were an agent with a limited mandate. Flag everything that relies on visual intuition, ambiguous copy or approvals outside the system.
Translated from the Spanish original with AI assistance and reviewed for accuracy. Read the original in Spanish.