Sierra Personal Agent Protocol: the money settles around the standard

The next contest in agentic commerce is not about which assistant is smartest. It is about the rules for how an agent is allowed to deal with a business, and whoever writes those rules is well placed to clip the ticket on every transaction that follows.

On 6 October, Sierra announced the Personal Agent Protocol, an open standard that Meta and Sierra are developing with Genesys, Instinct, Rocket, Shopify, Stripe and Walmart. It covers how a personal agent discovers a company, starts a session on a customer’s behalf, and signs in. The session is built on OAuth. The customer decides whether the agent gets read-only or write access, and the company decides what the agent can do through its website, its APIs or its own agent.

That is sensible design, and I would rather see an open standard than a dozen private ones. A direct connection beats an agent clicking through web forms.

But open standards seldom collect the toll directly. The money tends to settle at the points around the standard: who runs discovery, who sets the default permissions, who sees the session, and who carries the liability when an agent errs. A protocol can be free to adopt and still be expensive to rely on.

Sierra says more detailed permissions, push notifications and payments extensions could follow, with a v0.1 specification due later this month. Those are the parts where the money is. Whoever sits in the session when a payment is made, and whoever sets the default permissions, is in a strong position to charge for the privilege. Most customers will accept the recommended defaults, so the default effectively becomes the standard.

Two further risks are easy to miss. First, a business can lose margin without paying anyone. If agents compare offers across providers in seconds, brand, loyalty and upsell lose force, and the business risks becoming a fulfilment endpoint while the agent owns the customer relationship. Second, the party that sees sessions across many businesses learns demand patterns that no single business can see, which supports price discrimination and targeted advertising.

The announcement says a lot about access and visibility for companies. It says less about what happens when an agent’s interests and its customer’s interests pull apart, which is the question I raised earlier this week about how personal agents get paid. A protocol can say who is allowed in. It does not, by itself, say who answers when the agent gets it wrong.

Executives should treat this as a commercial decision, not a technical one. If your business will deal with customers’ agents, ask four questions:

1. Who sets the default terms, and can they change them without notice?
2. Is the standard open to challenge, and who governs it?
3. Can an uncertified agent, or one using a different payment rail, still reach you?
4. What will it cost you to be reachable?

Those answers will matter more than the protocol’s feature list.

#AgenticCommerce #AustralianBusiness #DigitalStrategy