Agentic commerce and control: Getting the balance right

An agentic commerce protocol is a set of technical and operational rules that define how AI agents are authorised to act on a shopper’s behalf, how payments are executed, and how intent, consent, data, and liability are handled in delegated transactions. The priority is not choosing a single protocol early, but making sure your payments and risk infrastructure can support multiple protocols over time.

Agentic commerce shifts discovery, decision-making, and (soon) the checkout into AI systems such as ChatGPT, Gemini, and other conversational interfaces.

These systems can surface products, compare options, and will soon be able to initiate transactions on a shopper’s behalf. That reduces direct interaction between you and your shopper.

Control depends on who defines intent, who authorizes the transaction, and who holds the proof of purchase. Businesses that retain ownership of mandate and payment infrastructure preserve visibility and accountability.

Ownership follows intent and authorization. If a third-party platform operating AI systems defines what an agent can do and holds the mandate, it can influence the customer lifecycle beyond the transaction itself.

To retain ownership, businesses should maintain merchant-owned mandates, auditable authorization records, and control over payment tokens and lifecycle data.

The primary risk is early lock-in to closed integrations, platform-specific tokens, and fragmented payment flows, which can be expensive to unwind. As protocols evolve, businesses need infrastructure that supports flexibility across AI platforms rather than committing to a single ecosystem.

Automation increases speed, and errors scale with it. Agent-led transactions require clear authorization boundaries, fraud systems that distinguish legitimate agents from malicious automation, and defined liability frameworks. Without strong guardrails, disputes and compliance risk can grow quickly.