A New Way for Businesses and Personal Agents to Work Together

Summary

Meta and Sierra are developing the Personal Agent Protocol, an open standard with Genesys, NiCE, Decagon, Rocket, Shopify, Stripe, and Walmart, to define how personal agents interact with businesses. Muse, Meta’s personal agent, runs a cloud browser/computer for each user and acts only on user requests. The core argument: when someone asks Muse to book, buy, reorder, or resolve support, they have already chosen the business. If a site blocks the agent, the customer experiences it as the business refusing them. As agent-mediated errands grow, more current and future customers will arrive this way, making blanket blocking a customer-acquisition risk.

Today Muse uses connectors or direct website browsing. It follows two norms: the 'one honest person' test (would a reasonable person do this manually at this scale?) and the 'if every agent did this' test (would the system still function if every user made this request?). It skips actions that violate these norms, explains limitations, and requires explicit user approval for sensitive actions like sign-in, reservations, or purchases.

The proposed path improves predictability and efficiency. Businesses could guide agents through preferred flows, structured data, connectors, APIs, or business agents. Agent-to-agent communication would reduce brittle browser automation, lower latency, and fewer errors. The protocol aims to handle secure user identity, business steering, optimized APIs/skills, and visibility into agent activity.

For ad ops and commerce leaders, takeaways: treat agent traffic as high-intent customers, not just bots; avoid blanket blocking; prepare agent-readable product, inventory, booking, support, and offer endpoints; ensure consent, authentication, and identity pass-through; plan measurement and attribution for agent-driven conversions; and engage early in open standards that may reshape discovery, checkout, and customer relationships across retail, payments, and CX. Because partners span retail, payments, and CX, adoption could quickly affect endpoint requirements, bot-management rules, and conversion tracking.

Analyst Note

What's notable here isn't the agent capability itself — it's the governance layer. Meta and Sierra are proposing shared rails for identity, authentication, and agent-to-agent negotiation alongside partners spanning payments (Stripe), commerce (Shopify, Walmart), and CX (Genesys, NiCE, Decagon). That's a commerce-and-service coalition, notably not an advertising one, and its scope — securely conveying user identity and letting businesses steer agents toward APIs and skills — sits adjacent to the identity plumbing ad ops has spent years rebuilding post-cookie.

The key implication for UA and monetization teams is funnel compression. If a personal agent handles research, comparison, and purchase without rendering pages, the session-level signals feeding attribution, audience segmentation, and retargeting windows thin out. The article's framing that behind every agent visit is a customer also surfaces a relationship question that overlaps with CRM and first-party audience ownership, since businesses explicitly worry about maintaining the customer relationship — the same asset retargeting depends on.

Worth watching: whether agent identity tokens become an accepted measurement signal, or whether ad platforms build parallel standards.