Most AI agents in the contact centre still behave like a very quick front desk. They pick up a query, work out what the customer wants, and either resolve it or pass it on. When the conversation ends, so does their involvement, and anything left open falls back to the customer or into a human agent's queue.
Cisco wants AI agents to stay with the problem. At WebexOne 2026 in Austin, the company introduced Dialog, an agentic harness for the Webex AI Agent platform that carries a persistent understanding of each customer across channels, teams and time, and keeps working on their behalf until an issue is resolved.
Speaking to CX Today at the event, Vinod Muthukrishnan, VP and GM of Webex Customer Experience, places Dialog as the third stage in a progression that has moved quickly.
The first wave of agents handled conversations. They were intended to find the intent, answer the question, get something done. Now we are taking a step function up, which is managing relationships: thousands of conversations, hundreds of journeys, but one relationship.
How Cisco Dialog Builds Relationship Context for Webex AI Agents
Muthukrishnan says Dialog brings four things together. It builds relationship context by combining enterprise and customer data, it handles how an agent is onboarded and optimised, it runs a self-learning loop on every conversation, and it powers long-horizon task execution by fleets of agents operating within enterprise policy. Underneath sits a state engine that can run several journeys in parallel, so a customer who starts on a missed-flight rebooking and then asks for a refund kicks off a second journey without the agent losing track of the first.
For the customer, the aim is that every touchpoint feels like a single ongoing exchange, whether they arrive by phone, message or in person.
There should be no wrong door. No matter where in the organisation you come in, no matter whether it's inbound or outbound, there's complete context preservation. It should feel like the same continuous conversation is in progress.
Webex AI Agents That Keep Working After the Conversation Ends
Cisco's on-stage demo used a missed flight to show what that looks like in practice. A customer with a medical emergency asks for a refund that policy does not allow. Rather than refusing or escalating under pressure, the agent gathers context, finds the exception process, and tells the customer they no longer need to chase it. It then opens an internal ticket and argues the case with an escalation specialist, citing the customer's status, lifetime value and a recent delayed flight, and even checks insurance details with the customer's own personal AI agent.
