Digital workers are hard at work at a bank near you

In the early days of automation and AI, vendors rarely talked about the possibility of intelligent machines replacing humans. Jobs were not going to be automated; instead, they were going to be augmented, with machines doing the boring stuff and humans freed up to do more creative and interesting work. Virtually every AI and automation system vendor has embraced this mantra. And it was mostly correct; AI and robotic process automation (RPA) typically automate tasks, not entire tasks, so at most any human job loss was marginal.

2022, however, is the year of the Great Resignation. Whether job departures are historically high or not is debatable, but the general perception is that human labor is scarce. As a result, there is less fear for now that AI and automation will present pink slips to humans.

One company, WorkFusion, decided the time was right to provide digital workers as a potential replacement for human workers, not just as a complement to them. The company was the first in 2015 to offer the combination of RPA and machine learning, known at the time as “intelligent automation”. It now offers a set of digital workers who use the same combination of technologies. The digital workers, which you can see here (they look a lot like humans!), are each focused on a particular set of tasks. WorkFusion maintains that they include full jobs (at least administrative and supervisory jobs), not just tasks, although I suspect this will vary from organization to organization. And WorkFusion says digital workers will need human review of their decisions, at least early on in their employment with a company. Tasks currently performed by the six digital workers (each of whom have been given suitably diverse names and photos) include:

  • Insurance underwriter
  • Bank Customer Identity Program Analyst
  • Banking Entity Sanctions Screening Analyst
  • Banking Transaction Screening Analyst
  • Banking Due Diligence Program Analyst
  • Customer Service Coordinator (email inquiries and portal)

These all sound like full jobs, although the WorkFusion website says the insurance underwriter can take on perhaps 70% of the human underwriter role. This matches my previous research on a human digital underwriter – the easy decisions are supported by an AI system, but the harder ones require an experienced human eye. The other jobs of the digital worker seem to require less human expertise.

The flexible and learning digital worker

Adam Famularo, the new CEO of WorkFusion, told me that digital workers can be owned or leased by the deal, and can come on-site (probably won’t catch COVID there!) or be cloud-based . He said workers would get smarter over time, the machine learning models within them would be recycled from data captured by each worker deployed to a customer. He also said that new digital workers could be created by customers and partners, and that WorkFusion would eventually operate a “digital worker exchange.”

Famularo spoke to a slew of clients when he took over, and he said none were worried about the job loss issue. Many spoke of the labor shortage they were experiencing. All of the clients, he told me, reported having human employees working alongside digital workers, or humans who were redeployed to other tasks. He believes that most of the tasks performed by digital workers so far are quite tedious and would not be missed by any human. He did say, however, that he hears a lot about digital working as a replacement for outsourcing.

Digital workers at Bank of Montreal

To understand how a client uses these digital workers, I spoke with Adam Schabes, who is the US Director of Anti-Money Laundering (AML) at Bank of Montreal (BMO). He confirmed that digital workers already support core anti-money laundering activities at BMO, including sanctions screening, checking whether customers are doing or doing business with sanctioned people or countries on various lists. provided by the government. A human employee still makes the final decision, but administrative research tasks are performed by the digital worker. The digital sanctions control officer, whom WorkFusion named Evelyn, visits external websites and various bank apps to find out who the bank should not do business with. “Our employees who previously did this type of work had investigative backgrounds, but you don’t need that background to go through lists — it’s an administrative exercise,” Schabes said.

BMO is also in the process of setting up Tara, a WorkFusion digital worker who performs transaction screening. Like sanction screening, it is an administrative task that a digital worker is well suited to. Content revealed by Tara may go directly into the Suspicious Activity Reports the bank must provide to the Financial Crimes Enforcement Network. The combination of the work of Evelyn, Tara, and other machine learning models that BMO has used have all helped reduce the number of false positive suspicious events that humans need to investigate by about 25%.

Overall, Schabes said, WorkFusion’s digital workers have done quite well. At first, he says, his staff were very apprehensive about job security. However, after seeing how the digital workers performed the relatively menial tasks, morale was boosted in the department. People started asking, “Can robots take on this task too?” Some requests had to be refused so as not to overburden them. Human workers were able to transition into a more customer-oriented and revenue-generating activity by being freed from tedious administrative work.

AML staff had actually started naming their RPA bots before the concept of the digital worker was introduced by WorkFusion – they called the first one “Becky”. Someone even put a cardboard box with a face on it and tinfoil arms on a desk to represent Becky. This isn’t the first time I’ve heard of anthropomorphized software robots; we humans seem to like the idea of ​​them once we get used to them.

Schabes concluded, “We’re looking to do more,” and I highly doubt he’s the only one with that attitude. Particularly in banking and insurance, there is a lot of back-office work that machines can do quite easily. First there was RPA, then there was intelligent automation, and now there are fully fledged digital workers. Right now, I don’t think most human employees have much to fear from the current generation of digital workers, but who knows what’s next?