CHARLOTTE, N.C. — KeyBank’s head of robotic process automation (RPA) hopes to expand his team by up to 30% this year for a total of 120 to 130 people with three more build-run squads, or teams of 10 developers responsible for coding (build) and servicing (run) of an application or bots.

“I’m creating a whole different pillar of process engineers now under my team,” Dominic Cugini, KeyBank’s chief information officer of service digitization, said last week at the Bank Automation Summit. “I want to get the team revving so we can get through at least 20 to 30 ideas a month.”

The Cleveland-based bank plans to create a group of process engineers to expedite the RPA team’s processing of automation ideas, Cugini said.

‘Never automate a bad process’

Process reviews play a critical role in the development of bots at KeyBank, Cugini told the BAS audience. Rather than simply identifying where to automate, Key thoroughly reviews how to optimize a process ahead of automation.

“Never automate a bad process,” Cugini said, explaining that one reason he moved the bank off a federated — or decentralized — mode of bot deployment was because it’s more challenging to ensure that processes are optimized when development is decentralized.

Cugini should know: When he took over the RPA program in 2019, it had already been a federated model since mid-2017, with more than 100 bot developers.

“At the end of that year-and-a-half journey, we had less than 100 tasks actually in production,” he said. “We had $0 of hard savings or revenue that we could quantify.”

Cugini scaled down the RPA department and retained 13 of the original 100 bot developers, one manager and a couple of process engineers, after bank senior management gave him six months to determine whether RPA was worth pursuing.

“We want to have a value proposition for everything we did, so our goal was 18 months or less — we need a return on that investment,” he said.

Using a centralized deployment

Fraud disputes proved to be the first place Cugini and his team began deploying RPA.

Typically, fraud investigators pull information from multiple systems and spend anywhere from one to two hours building a case, plus an additional hour or two for review. KeyBank’s RPA team began using bots to collect documents from multiple systems so that the cases were ready for review by investigators when they arrived each morning. That system is still in place at Key today.

“Now we build those cases 24/7,” Cugini said. “So when people come in between 6 and 8 a.m., the first thing they do is go to the queue, pull a case and start working a case.”

The bots allowed the bank to reduce the cycle time — and more than double its capacity — for fraud investigation without adding a single person, he added.

“I was able to go to our CIO and CFO and say, ‘I think RPA is real’ and we showed some very good examples,” Cugini said.

Using a centralized approach that reviewed processes ahead of automation, he took bot deployment from just over 100 tasks — retaining only two of the original bots — to more than 600 tasks automated. The team also doubled in size that year.

Build-run approach

KeyBank divides its developers into squads of 10 and uses a build-run approach so developers are also responsible for servicing the apps or bots they build, rather than the traditional approach, which separates those functions. The bank uses both Automation Anywhere and UiPath RPA solutions as well as agile development, which allow squads to deploy quickly and then iterate as the business provides feedback on the demos.

“If we have a conversation on Monday, our goal is by Friday or the following Monday that we at least have a draft and a semi-working demo and we can continue to iterate faster,” Cugini said.

This team-based approach allows Key’s squads to build at around 30% less than the average bot costs, Cugini estimated.

A squad might be divided into sub-squads and work on a maximum of three projects at a time. This creates a good manager-to-team ratio, he added.

Build-and-service tickets are tracked to try to establish a full-time equivalence ratio. As that ratio starts to grow, it can trigger a need to hire more teams.

The squads also focus on reuse of code. Engineers at the bank hold coffee chats twice a month to share best practices and what they’re developing. Managers also host team town halls to provide developers an opportunity to share what they’re working on.

“I would say there’s probably 25%-plus times savings in that, because you have the core code, you don’t have to remodel or rethink and re-engineer how that works,” Cugini said.

Big Idea Squad

Business units at KeyBank can submit ideas, but ultimately it’s up to Cugini’s Big Idea Squad to decide which processes to pursue.

For instance, the team might look at anti-money laundering (AML) as everything that happens with AML, including disputes, all channels, all processes and technologies, he explained.

“Yes, many of our processes are very, very complex. Many of them we’ve made very, very complex,” Cugini said. “That’s where good process engineers that can simplify those and take an eight-step process into a three-step process, which sounds counterintuitive.”

The Big Idea Squad reviews typically take eight to 12 weeks, at which time the team will have developed a full business case, he said, adding that they will share the proposed plan from both a technology and business perspective, as well as the value proposition for the changes.

What’s next

The RPA team has enough work to keep it busy for the next two years, Cugini said. Among its top priorities for this year: an end-to-end assessment of the fraud space.

“We really want to secure our fraud area for our clients and our shareholders,” he said, adding Key will be looking at governance, risk, and compliance and first-line controls. “At the end of the day, the goal is we automate 100% of those — because if we do that, we’re going to be much more secure for risk posture than we were before we started.”