NAB has 250 – and sometimes as many as 500-plus – “virtual workers” or bots employed on a range of banking processes, courtesy of a five-year effort to seed “connected automation” across the enterprise.
Speaking at last month’s SS&C Blue Prism World On Tour event in Singapore, head of technology for workflow automation and robotics Pamela Stooke and manager, engineer (robotics) Vikram Thapar provided the first real view of NAB’s efforts since they were first briefly announced back in 2017 .
For observers familiar with NAB’s cloud journey, there are some striking similarities in the approach and particularly the “industrialisation” of the technology within the bank, insofar as it was a series of guardrails, governance measures and factory-like repeatable actions that really helped set foundations that earned and engendered internal trust in the cloud, paving the way to achieve scale.
The path to adoption of robotic process automation (RPA) at scale is very much similar.
Thapar and Stooke recalled RPA’s beginnings at the bank.
“It was actually born on laptops in a room in one of our buildings,” Thapar said.
“One day, one of my head office [team] told me, ‘Let’s visit the automation room’. I entered that room and saw 30 laptops and people juggling and scheduling bots.”
A lot has happened in the five years since, but the consensus is that NAB has come a long way from “a laptop journey to a full-fledged factory to deliver tangible outcomes – not only from a cost efficiency perspective or productivity, but the right controls, and having that mindset within the business of, ‘Let’s go, let’s let the automation do the work, we trust the bots to do what they’re told to’.”
The bank’s first forays on RPA did not reach the heights that NAB had set for the technology.
“It started off in 2017, we got some bots running, but what we delivered was not fit for purpose and the ROI wasn’t what we wanted,” Stooke said.
“We said, ‘It’s not the technology, it’s how we’ve implemented it. Let’s have a look at what governance is lacking.
Thapar said the bank’s executives and senior leadership still “believed in the technology” at this point, and it was agreed to give RPA another go: “Let’s do this properly,” Thapar said.
This time, the bank wrapped governance mechanisms around the technology adoption that Thakar indicated had proven to be of value.
This included “live proofing”.
“You can’t get everything right in non-prod – you don’t have all your test data set up, so we came up with a live proofing concept,” Thapar said.
“I personally call it a probation period of a bot in production, where you kind of hand-hold the bot, and make sure you’re getting through all the test cases and scenarios.”
The team also focused on technology change governance for its target applications. This was about how to adapt and retrain bots in instances where the target applications that the bots relied upon changed.
NAB currently has a portfolio of “50-plus target applications” for which it has operational patterns that pertain to bot use and interaction.
Thakar said the number of applications that can interact with bots continues to grow. He said the list is a “key success driver for us to say ‘yay’ or ‘nay’ quickly [to a bot use case request] based on the target application [it would need to work with].”
Governance efforts were also focused on building the confidence of internal teams and business units, as well as of auditors and risk partners, to know whether a bot could perform certain sensitive transaction types.
In addition to governance, the technology became more scalable when it was rehosted on NAB’s Azure-based private cloud.
As part of that move, NAB also set Blue Prism up with a blue-green deployment model.
“We’ve always got two versions of Blue Prism, which has been a significant enabler for us to tackle upgrades, and with balancing key features between the two applications as well,” Thapar said.
All of this was completed between 2017 and 2019, and that left NAB well-placed for what would prove to be a breakout moment for the use of the technology: the onset of the pandemic.
Stooke said the immediate demand for bot-based services came from business units that already used the technology.
“Time and pressure on our space was just starting,” Stooke said.
“Where it came first was where we have engaged operational stakeholders with automations already in place.
“They knew what the bots were capable of. They were first asking we need to scale up – easily done.”
From there, the connected automation function at NAB had progressively less time to deliver automated processes, such as to help the bank triage hardship and loan deferral applications.
Thakar noted it was uncharted territory for the function, having to stand up capabilities in as little as half a day, and having to coordinate all work remotely via Zoom.
Still, the efforts kept pace with internal demand, and Thapar noted the bank still has a “healthy backlog” of potential automation use cases.
One of the newer uses is handling a large-scale data migration for UBank that would otherwise have required up to 100 full-time equivalent (FTE) staff to run.
Thapar also pointed to the continuation of the “industrialisation” of connected automation at NAB.
“We look at an automation end-to-end, where we build, design, deliver and clear proof for a bot within three months, and give the bank ROI [in] the same year, if not earlier,” he said.
He added that further improvements are being made to make it more transparent what work bots are doing and how their actions might be improved.
“We’re putting together a real-time Power BI dashboard which would enable the business to have real-time diagnostics of what the bot’s doing, the error rate, and what needs to be reinvested,” he said.
“There’s a lot of continuous improvements as well which we’re targeting.”