Say goodbye to your bank teller and your insurance sales agent. Dozens of financial services jobs like these are starting to be replaced by technology. Over the next 15 years, AI will take even more of these types of jobs, according to research firm Autonomous’s new report.
Two converging trends have enabled forms of AI that can effectively mimic or replace human labor. On the one hand, specialized hardware has increased processing power, making it possible for AI systems to generate outputs in real time. At the same time, the amount of data available to feed those systems has skyrocketed, thanks to search histories, online photos, and more. Combine those two capabilities, and software engineers have the tools to create virtual assistants like Alexa, and automated filing services like Google Photos.
So far, AI is best suited to discrete, repeatable tasks. But over time, forecasters say, it will be able to manage more complex assignments. Already, AI is better at image recognition than humans: In a 2015 competition, a team of ResNet researchers achieved an image recognition error rate of just 3.57%.
“We are on the verge of software gaining competency to perform service work,” says Lex Sokolin, global director of fintech strategy for Autonomous. “The key thing that AI cannot do yet is context switching, which implies that a chess robot cannot drive a car, and a chatbot cannot see.”
In financial services, AI’s impact on jobs is likely to be particularly dire for certain categories of employees. Sokolin projects that by 2030, AI implementation could save U.S.-based financial services companies $1 trillion through productivity gains and reduced head count.
Banking and lending could see the largest change, he says, with 1.2 million jobs at risk and a potential $450 billion in savings. Insurance follows, with 865,000 jobs at risk and a projected $400 billion in savings. Lastly, there are 460,000 jobs at risk in the investment management sector, equivalent to as much as $200 billion in savings.
Such shifts will not happen overnight. In Sokolin’s view, AI will drive gradual change over the next seven years, and then accelerated change between 2025 and 2030.
“Several trends in AI are increasing quickly—the evolution of hardware, open source software, academic papers in the field, and venture funding. But in other areas we are likely to hit limits,” he says. “We think that it will take a meaningful time for AI to become culturally ‘normal,’ which is why we see large impact in the latter part of the next decade.”