AXA's admin bots save $250k in six months
- 06 February, 2019 06:01
The UK arm of insurance giant AXA has deployed 13 software bots across its business since last summer, aiming to help employees with mundane, repetitive admin tasks like filing customer correspondence in the claims department, saving 18,000 people hours, which equates to roughly £140,000 (AUD$250,000), in productivity gains since.
"Back in 2018 we were seeing more and more discussion and demand over improving productivity internally and boosting the consistency across customer facing parts of our business with claims, sales and service," Simon Clayden, chief operating officer (technology) at AXA UK told Computerworld UK.
As a result the company started to look at robotic process automation (RPA), a burgeoning technology area where software 'bots' can be programmed to complete simple, process-driven tasks on a user's computer.
The primary aim of the project was to help free up staff from mundane tasks, it was "never about robotics looking sexy, it was about productivity and improving things for our people," he added.
After a vendor comparison of three specialists - UiPath, Blue Prism and Automation Anywhere - AXA opted for the New York-based UiPath, thanks to its breadth of features and the quality of its support and account management, Clayden said.
When Harry met AXA
The first bot, called Harry, was rolled out to staff in the customer property claims team in June 2018. It was soon joined by Bert in the commercial property team and Lenny in the liability team. All the bots were named by employees of those departments.
Broadly speaking all of these deployments are cases of 'attended' RPA, where the bot helps employees get a task done, rather than taking it completely off their hands. Most of the bots to date focus on 'reading' customer correspondence and matching them with the relevant claims records. Where this would take a human on average of four minutes, it takes a bot 42 seconds, the company claims.
As Clayden explained: "When a customer writes to us or emails us, that piece of correspondence would be scanned and enter a queue and the employee would do something with it and move it from scanning to the claims system. It was essentially dragging from one workflow to the claims system, there is a clear business logic of what needs to be done, which made it a prime candidate [for RPA]."
AXA also embarked on a concerted communications effort to dispel any concerns that these bots would be replacing human workers. "Very quickly we saw positive feedback and operational teams were writing their own blogs and taking ownership and saw the robots taking those repetitive tasks away from them," Clayden said.
Word soon got out of these admin-busting bots, and other departments started asking for "their own 'Harry 2' to arrive," he added.
When stacked up against the cost of implementing the technology and standing up the underlying cloud environment in Microsoft Azure, Clayden said the margin of savings is still significant, while also stressing that any savings "are nothing to do with replacing people, it is a productivity improvement."
"They provide on average a 2x return on investment per robot, that's our benchmark," he added.
RPA vendors like UiPath increasingly look to differentiate themselves by talking up their artificial intelligence capabilities, promising bots that will learn and adapt as they go, like a human employee would.
Clayden doesn't see this becoming a near term reality however. He says the robots right now are "like toddlers, they're not even pre-school age right now. They need to be taught a lot and require a lot of looking after.
"They are good at doing one thing, but their ability to make decisions is quite far off."
Short vs long term gains
Not everyone is delighted with their new robot coworkers. Some members of the IT team have shown their distaste for RPA being brought in to patch over more pressing long-term issues with the technology stack across the business.
"Questions tend to come from IT people that ask if this is a long term solution or a plaster over the cracks involved in not investing in ageing technology and diverting that resource to replacing our ten year old policy administration system, for example. They are concerned by that," Clayden said.
"We don't build a robot to mask technical debt and still invest in the core platform as much as the robots," he added.
That being said, staff within the monitoring function of the IT team are already eyeing RPA use cases to better surface alerts and automate elements of testing and app monitoring be able to see and react faster to problems at the infrastructure level.
The rest of this year is really just about scaling out the use of these bots and developing the skills at AXA to do that.
"We now have a sizeable order book to work through and have to deliver on that aggressive plan for this year to add 50-100 new robots," Clayden said.
The main barrier for this land and expand strategy will continue to be skills however. "We are finding a shortage of skills out there," he said. "We knew it was quite a new technology, but didn't realise how scarce the quality of that resource is."