Telstra is using agentic AI to interrogate its customer data in a bid to ease pressure on its frontline call centre operations.
The carrier revealed that it has started using the technology to identify errors in customer data after they’re migrated from legacy relationship management systems, allowing it to act before customers reach out to call centre staff.
Speaking at Microsoft’s AI tour in Sydney, Telstra AI solutions group owner Kim Bennemann said that the capability was one of the most valuable use cases the carrier had found for agentic AI to date.
“That would otherwise be handled reactively, so waiting until a customer calls. It becomes a really, really operational heavy burden for our front lines, so that’s a real initial [value] that we’re focusing on at the moment,” Bennemann said.
Bennemann said that the customer migrations typically happen in three phases, the first being to identify those that need to be moved and the second being the move. The third she said was “where the fun starts” and “discrepancies” start to appear.
“We’re targeting in on that third phase, once the customer is actually migrated, using Microsoft AI Foundry agentic capability to build agents that can reason over that data to help us identify customer issues,” Bennemann said.
“Obviously from a benefits perspective we’re going to reduce the calls coming into the call centre, we’re going to create a better customer experience.
“And we’re actually going to reduce the [time] of these migration projects as well, because we can actually migrate these customers a lot quicker than what we could do before having agentic capability in place,” she added.
The agent was expected to be used accelerate other Telstra system modernisation projects, Bennemann said.
“Traditional migration projects generally [require] a custom built, very specific solution for that particular migration, but how we’re thinking about it now and how we built out this, the agentic solution is actually smaller decoupled components.
“The agents reason over this data and perform various different functions [that are] not specific for these use cases,” she said.
The technology had also led to a broader shift in Telstra’s strategy for managing its data. Like the agents created for it, the project led to the creation of data products – reusable data sets – that the carrier could also use for multiple customer migrations.
Bennemann conceded that, early on, this was an early hurdle that initially stymied the migration project.
“There was a little bit of a slow start because of this and we had to actually work through find the data but what we actually did in that process was rather than create very bespoke methods of retrieving that data and bedding it into the solution is we actually created data products as well.
“What that means is [that] while this project did a lot of heavy lifting in terms of the specific data products needed for this … for any AI project that has similar data requirements, rather than have to reinvent the wheel every single time, they can actually leverage these data products.”
