One of the fundamental questions leaders in the banking world have looked to answer over the past year is whether artificial intelligence will replace human jobs.
Views on the matter vary wildly.
Some anticipate the impact of AI on banking jobs will be akin to the effects of automated teller machines on tellers. That's the view of Zor Gorelov, CEO of Kasisto, a conversational AI company that builds products for the finance and banking sector.
"Look at the ATM — it didn't replace tellers, it just took over some of the more simple tasks, but many people still count on bankers to handle some of the more complex activities within the branch," Gorelov said. "AI is doing the same across a number of functions within the banking ecosystem."
Others have a more dire view — namely, that AI will eliminate jobs. That is how Sam Altman, CEO of ChatGPT maker OpenAI, sees the situation.
"A lot of people working on AI pretend that it's only going to be good; it's only going to be a supplement; no one is ever going to be replaced," Altman told The Atlantic in an interview last year. "Jobs are definitely going to go away, full stop."
The changes to come
Most professionals in the finance industry lean toward Gorelov's view, according to a survey of more than 300 leaders and staff at banks, fintechs and related companies, conducted in October by Arizent, American Banker's parent company. In the survey, 75% of respondents said they believed AI would change the nature of some jobs but wouldn't replace human workers.
Twenty-one percent of respondents agreed with Altman — that AI will replace jobs in the banking industry. Only 4% said the tech will have little impact on banking jobs.
The data reflects the views of Gorelov and those who agree with him. There were roughly 6.7 million finance and insurance jobs as of December, according to Federal Reserve Economic Data. That's up almost 900,000 jobs, or more than 15%, since December 2000. For comparison, the overall number of employees has increased by roughly 18% over that time period, according to FRED data.
Over the last two decades, there have been numerous technological advancements that have irreversibly changed the industry — wider adoption of the internet by consumers and businesses, the rise of online then mobile banking, person-to-person payments and, most recently, real-time payments.
Many of these changes have made banks more efficient and eliminated mundane tasks but didn't necessarily lead to overall job losses. Instead, financial institutions have added to their headcounts in other functions, or instead of eliminating jobs entirely, they have expanded certain workers' duties to cover new responsibilities once their time was freed up from doing more routine tasks. The same could prove to be true with wider adoption of AI.
"Will we replace people? Probably not," said David Becker, CEO of First Internet Bank in Fishers, Indiana. "Can [our employees] do other things, and can we grow without adding people? I think that's where the real power comes in."
The areas most ripe for disruption from AI are currently being considered. In the October research from Arizent, about a quarter of respondents cited customer service roles as the top use case for AI. Fraud prevention and detection, analytics and marketing were also high on the list.
Becker believes that for the most part, rank-and-file employees in financial institutions do not stand to lose their jobs to AI. Rather, it is people in middle management and some upper management that might be more at risk of losing their opportunities.
Middle management faces a greater risk of replacement, Becker said, because of the amount of work people in those roles put into obtaining and processing documents — financial statements from customers, tax returns from the IRS and so on. The data entry part of this work — putting key figures from those documents into spreadsheets — is where AI can best help with automation.
Many experts also expect that back office jobs — those involving data entry, maintaining records, executing trades and others — are easily augmented, though not necessarily replaced, by AI. The technology will help these workers increase their productivity but will also require that they learn new skills to properly utilize it, said Giles Wrench, vice chairman of the consulting group JLL Financial Services.
This need for upskilling means banks will need to create roles focused on training employees in data fluency and analytics, both of which will be important to take advantage of new AI technologies, Wrench said.
"Even in these roles, AI integration will likely augment human capabilities, requiring professionals to advance their skills to handle more complex activities rather than completely replace jobs," Wrench said. "To successfully deploy AI, banks will need to invest in upskilling their employees to train them on new technologies that can supplement their jobs."
Valuable new skills
In 2023, ChatGPT and other large language models — Anthropic's Claude, Cohere's Command and Meta's Llama, to name a few — took up a lot of the oxygen in the room, and for good reason. The capabilities these models exhibited took many by surprise with their ability to wield language in ways that seemed human. But language is only one area where AI has developed expertise.
Prior to the recent advent of language models, many banks had already mastered the use of artificial intelligence in the detection of fraud, money laundering, financing of terrorism and other financial crimes.
These functions can be helped by AI applications because of their data-intensive nature, but a human is still needed to oversee the task. As AI has become more ubiquitous in various bank departments, institutions have created new jobs to oversee the technology.
This applies even to small institutions; Michigan State University Federal Credit Union has created a role to manage and perform training on Fran, the credit union's conversational AI, according to Ben Maxim, chief digital strategy and innovation officer.
AI will also create the need for higher-level roles, according to Rajesh Iyer, global head of machine learning and generative AI for financial services at the consulting firm Capgemini. These roles may include AI solutions architects who design AI-based solutions to meet specific banking needs and AI ethics officers who ensure the institution's use of AI aligns not only with regulations but also ethical guidelines.
Increasingly capable AI is likely to shift bankers' roles away from rote and mechanical jobs toward more analytical ones, allowing a person who isn't a spreadsheet wizard to achieve a greater level of analytical abilities.
"The people in the accounting department and our CFO — they can do things with spreadsheets I can only dream about," Becker said. "But when I can just ask simple questions and pose what-ifs in simple language, and instantly have responses — I mean, that's a phenomenally powerful tool."
One of the biggest things the $5.2 billion-asset First Internet Bank will do in 2024, according to Becker, is figure out how to make effective AI tools to simplify the day-to-day work of employees. First Internet Bank already uses AI for transaction monitoring — fraud and financial crime detection included — but besides that, Becker said the bank hasn't even scratched the surface of what the technology can do.
"I think AI at the end of the day will be more impactful than the internet," Becker said. "The internet allows you to accumulate data; it allows you to spread data at a national level very effectively.
"But it doesn't really allow you to manipulate data or crunch the data," he added.
AI, on the other hand, does.
This also presents risks, Becker acknowledged, particularly around data security. A model trained on a vast amount of proprietary data might serve as a tempting target for hackers, but Becker also worries about proprietary data ending up in a public dataset as a result of employees trying to train a model to help with analysis.
His fears have some foundation in reality; researchers have found ways of tricking ChatGPT into providing real contact information for real people, apparently derived from the system's training data, simply by asking the chatbot to repeat a word over and over again.
In addition, customers hoping to use products by OpenAI and others that promise enterprise knowledge management — a chatbot alternative to searching vast troves of documents for niche information — often must provide these companies access to the knowledge they are looking to manage. While these companies promise security of the data, banks are left to assess these vendors' data security practices to ensure they don't suffer a third-party breach.
"The point that I've really tried to drive home to people who just get excited about a salesman talking about all the great things they can do is just make sure you have the proper security up front," Becker said.
"You're truly giving employees the keys to the family jewels, and they can blow it up very quickly — inadvertently, but very quickly — and it's really nonrecoverable," he added.
