The big automation trends for 2021 are not surprising; robotic processing automation (RPA), artificial intelligence (AI), machine learning and natural language processing all top the list. What’s interesting, though, is how these technologies will combine in 2021 to create new opportunities — and challenges — on the path to automation and digitalization.

Trend No. 1: RPA

RPA will continue to be a hot technology in 2021, with revenues expected to increase 19.5% over 2020 levels, according to the IT research firm Gartner. In fact, the RPA market is a safe bet for the next four years, with double-digit growth predicted through 2024. As a cost-efficient and fast way to introduce automation into tech stacks, many financial institutions turned to RPA to meet volume demand for the Paycheck Protection Program (PPP).

“RPA is used by a number of organizations — many of which had never used it — to stand up PPP,” Nicole Sturgill, vice president at Gartner Banking and Investment Services, told Bank Innovation. “It’s a faster way of handling automation than IT, which has whole system conversions they’re working on.”

The challenge: Fragile systems

When lines of business invest in a solution without including IT, complications are created around running multiple solutions with no centralized oversight. While those solutions may be best of breed for what they’re designed to do, when IT updates aren’t synchronized with the automation management team, problems can arise.

“If you update a system that RPA is pulling from, the next time it queries, it won’t work,” Sturgill said. “If anything changes on any side of the system, if the person responsible for that bit of code does not know about that system change, then there’s no way to fix it. It will break.”

IT can address the problem by centralizing the oversight of RPA solutions, even if multiple solutions are running. This ensures IT knows all the integration points and can support them during system upgrades. In the long run, if the process is subject to frequent changes, IT may want to replace RPA with an API-based solution, which allows for more flexible integration, Sturgill said.

Trend No. 2: AI, natural language processing and machine learning

Banks and credit unions are investing in AI and similar technologies to automate and improve more functions, Jim Marous, CEO of the Digital Banking Report, wrote in a November report, “Innovations in Retail Banking.”

“One of the most common uses of automation and AI is in customer service, where firms use data, analytics and automated systems to respond to basic inquiries from consumers,” Marous wrote. “This not only saves money, but improves the standardization of solutions, allowing humans to be used for more important tasks.”

Gartner foresees the next step is for organizations to combine AI, natural language processing and machine learning with other technologies to create what Sturgill calls “hyper automation.”

“A very simple way to think about hyper-automation is automation technology to automate across the process,” Sturgill said. “It’s that combination of technology.”

So, for instance, loan data may originate with optical character recognition and RPA may be used to move the data, while AI may be used to crunch the data from an RPA system. AI can also be combined with chatbots, another automation technology expected to grow in 2021. By adding AI, natural language processing or machine learning technologies to other automation tools, organizations can create more complex systems that better meet business needs.

To that end, University of Pennsylvania’s Wharton School of business suggests organizations think of AI as providing a supporting role in “the broader strategy of the company,” rather than thinking of AI and similar technologies as stand-alone solutions.

The challenge: Automating bad processes

“The challenge of hyper-automation isn’t to pick just a technology but to identify the best process for customers and employees,” Sturgill said. This is already a challenge with some installations of RPA.

“Many organizations buy it and it works well, then organizations think, where else can I do this? So they try plugging it in. A process might begin bad, so now you’ve automated one step in a bad process,” she said.

Sturgill suggested a better approach: Figure out the business problem you’re trying to solve, then, if necessary, redesign the process to ensure you’re automating the process you want, not the process you already have.