Wonderful agents can now work with your customers’ agents

Starting today, Agent Access is open to design partners on Wonderful: businesses can recognize personal agents acting for their customers, decide what each one may do, and let approved work run across their workflows.

Companies spent thirty years teaching their contact centers how to handle people: queues, identity checks, scripts, escalation paths. Personal agents arrived in a single quarter. They do not wait, they do not tire, and they do not take no for an answer. They call at 3 a.m., fail the identity check, ask for a human and try again at 4. Every minute is billed. The rules of engagement were written for people. The agents did not read them.

Today an agent acting for a customer has to pretend to be one. It waits on hold. It clicks through pages built for human eyes. It does slowly, and at real cost to both sides, what a direct channel does in seconds. Without a way to know who the agent is and what the customer approved, a business has two bad options: keep every agent in the slow lane, or open a fast lane it cannot control.

We are adding a third. Agent Access has three parts.

Detection. Wonderful flags when a likely agent is on a call, in a chat or on the site, and routes it to the agent path instead of treating it as a person.

Policy. The business sets the rules once: which agents it recognizes, which actions they may take, which steps still need the customer, and which are never allowed. Permissions expire, and the customer can take them back at any time. If a task needs more than was approved, the agent asks the customer in the moment and continues, without starting over. The agreed rules decide, every time.

Execution. Approved work runs beyond the service conversation, across the workflows the business already runs on Wonderful, with a record of who acted, for whom, and under which approval.

Agents that can prove who they are and what the customer approved get the direct channel. Agents act only within that approval; nothing beyond it is available to them. And agents that cannot present a credential yet, which is most of them today, get an on-ramp: the customer signs in and approves what the agent may do, and the agent gets a short-lived pass to work within those rules until it adopts a shared standard.

All of it is built on open standards. Standards for agent identity and consent are taking shape across the industry, backed by its largest names, and we are building on them together with others, because no single company should define this alone.

It is also the first step toward something bigger, because the relationship runs both ways. It is 3 a.m. and a card is flagged in Lisbon. The bank’s agent reaches the customer’s agent under the agreement they already hold: confirm or block, nothing else. The customer’s agent knows she landed in Lisbon six hours ago, and confirms. The contact center never sees the case. That is not access control. It is an agreement between two agents, written by both sides: when each may contact the other, which actions are allowed, which guardrails apply, and when it ends. The customer decides what to grant. The business decides what to allow. That is what we are building next.

Agent Access is open to design partners now. The next posts in this series will cover how an agent earns the fast lane, and how the shared agreement is set and kept. If you are working on personal agents, on either side, or your contact center is already hearing from them, talk to us.