AMP’s big bet on bots
Originally posted by Rohan Pearce @ computerworld.com.au
AMP has rolled out what Frank Cammaroto likes to describe as an “ecosystem” of bots that are responsible for receiving, delegating and acting on an increasing proportion of service requests at the ASX-listed financial services firm.
Cammaroto, AMP’s acting director for technology and operations services, yesterday told the Quest Ignite AI & Machine Learning Summit in Sydney that the company had rolled out a range of bots based on Wipro’s Holmes AI and automation platform.
The automation push really came to life when AMP centralised its technology operations.
“That gave us an opportunity to form a function that was focused on operational excellence and that had at its heart automation and innovation to deliver on our services,” he said.
AMP looked at four categories of automation it could deploy: Ranging from process automation through to infrastructure automation, monitoring automation and AI/cognitive automation.
“Our approach to implementation at AMP was through a maturity model,” Cammaroto said. “The way we approached it was to identify how we can go from simple script-based automation through to process automation and then on to some cognitive style automation.”
The company’s “sweet spot” turned out to be enhanced process automation and supervised machine based learning, he said.
Currently some 15 per cent of overall service requests are handled by the bot ecosystem
A bot equipped with natural language processing capabilities — dubbed Chat A Lot Bot — is a key channel for service requests, alongside the employee contact centre, online portal, email, and automated alerts.
A delegator bot then uses supervised machine learning to help route some 600 tickets a month with an accuracy of 73 per cent.
Two ‘actioning’ bots are able to automatically deal with more than 1000 tickets a month through API calls or through command line integration, with a cycle time reduction from 7.5 hours to 5.6 minutes. Some 60 per cent of system access support requests are currently handled by the bots — and Cammaroto’s team is “going to try to hit 100 as fast as we can”.
“The level of automation and the accuracy of automation continues to improve and will increase over time,” he said.
An additional bot combs through data looking for insights that AMP can leverage.
Although “each bot in isolation can deliver some amazing value,” the real power is when they are combined into an ecosystem, Cammaroto said. “Through the ecosystem we’re able to then start influencing directly the end-to-end service experience of our customers and the services that we deliver,” he said.