Share with your CIO
Sierra and Meta are betting that consumer AI agents, software that shops, books and negotiates on a user’s behalf, will become a primary entry point into enterprise systems, and they want to define the protocol before anyone else does. The Personal Agent Protocol uses OAuth for session identity, lets consumers grant read-only or write access, and routes agent traffic through a company’s website, MCP or OpenAPI interfaces, or the company’s own AI agent. Genesys, Stripe, Shopify, Walmart and Rocket Companies are named partners. A v0.1 spec is due later this month.
What this means for your business
The dividing line this protocol draws is between enterprises whose APIs and identity infrastructure can handle agent-initiated sessions today and those that can’t. Write access, the ability for a consumer’s agent to change an order or update an account without a human in the loop, is not a theoretical future state under this design; it’s the default second option at sign-in. If your API governance hasn’t been stress-tested against autonomous actors with delegated credentials, that gap becomes a live exposure the moment any meaningful share of your customers adopts a personal agent.
The competitive tension buried inside the announcement deserves more attention than it’s getting. Sierra sells AI customer service agents. Genesys sells its own Agentic Virtual Agents and now runs Sierra agents inside Genesys Cloud. The protocol lets a brand choose which agent “answers the door” when a consumer’s personal agent arrives, and Sierra says anyone can implement the spec. But the v0.1 draft will reveal whether the routing logic is genuinely neutral or whether it subtly advantages Sierra’s own agent-to-agent pathway. Enterprises that are already contracted with Genesys, Salesforce Agentforce, or ServiceNow should watch that routing specification closely before it hardens into a de facto standard.
The Gartner figure Genesys cited, that 80% of organizations will have AI agents consuming the majority of their APIs by 2028, is the right frame for the urgency here. Standards races tend to close faster than enterprises plan for, and the companies joining design workshops now will have more influence over how brand controls, permission scoping, and liability boundaries get written into the spec than the companies that wait for general availability. The decision this reframes isn’t whether to support personal agents eventually; it’s whether your current API inventory and identity architecture are the ones you’d want enshrined in an open standard.
Concept deep-dive: OAuth in agent sessions
OAuth is the authorization framework most enterprises already use to let third-party apps act on a user’s behalf without sharing the user’s password, think “Sign in with Google.” The Personal Agent Protocol applies the same mechanism to AI agents, so a consumer’s personal agent presents a scoped credential rather than raw login details. The business implication is that your existing identity and access management stack becomes the first line of control over what any external agent can do inside your systems.
Based on reporting from Genesys Joins Sierra, Meta on Open Standard for Personal AI Agents, originally published 2026-10-06 18:31:00.
