Citi Utilizes Generative AI to Analyze 1,089 Pages of Capital Regulations

As Wall Street increasingly adopts the emerging technology, Citigroup Inc. intends to provide the majority of its 40,000 coders with access to generative artificial intelligence.

As part of a small pilot program, the Wall Street giant has quietly allowed about 250 developers to try generative artificial intelligence, a technique favored by ChatGPT. Now it plans to expand the program to most of its coders next year.

The bank and its rivals have slowly begun experimenting with the technology, which made waves last year when ChatGPT debuted and showed how generative AI could produce sentences, essays or poetry based on a user’s simple questions or commands. The technique typically creates this new work after being trained on vast amounts of pre-existing material.

Bank executives are increasingly claiming that AI will make their employees more efficient. Like when federal regulators dropped 1,089 pages of new capital rules on the US banking sector, Citigroup went through the document word for word using generative AI.

The bank’s risk and compliance team used technology to assess the impact of the plans, which determines how much capital the lender needs to set aside to protect against future losses. Generative AI organized the proposal into parts and drafted key guarantees that the team presented to outgoing treasurer Mike Verdesch.

At Citigroup, the arrival of ChatGPT awakened the use of artificial intelligence in joint work. In response, the bank formed two task forces earlier this year to explore potential uses for the technology, according to Stuart Riley, the bank’s chief information officer, who is overseeing the effort.

“This affects every part of the bank,” Riley said in an interview. “Some of them are small, helping with the daily routine, and others are complexlegalamost-off.

Bank employees have been increasingly concerned that technology might replace them. According to Riley, this is not the case. Whether AI or workers create a line of code, it still needs human oversight, he said.

“People are still looking at the code to make sure it’s doing what they expected it to do. They’re still monitoring like a coxswain,” he said. “The AI tool is given to the developer so they can produce code faster – it doesn’t replace them. We’re using artificial intelligence to empower our employees.”

JPMorgan CEO Jamie Dimon echoed a similar sentiment this month when he said AI is likely to make dramatic improvements to the quality of life for workers, including reducing the workweek for some to three and a half days. The technology is already being used by thousands at his bank, he said, and between February and April the lender advertised for more than 3,500 related roles, according to consultancy Evident.

Citigroup is also looking into modernizing its systems with the help of artificial intelligence. According to Riley, this process would normally cost millions of dollars and require significant manpower. To update legacy systems, the banking giant needs to change its coding language, and AI can help translate it from an older one like Cobol to a more modern one like Java.

The bank is investigating ways to use generative artificial intelligence to analyze documentation and create content. AI can speed up the process of parsing quarterly results by analyzing earnings and other public documents, freeing up staff to spend more time with customers instead of crunching numbers.

Improving accuracy

Checking datasets for errors and anomalies and improving data reconciliation are other use cases the bank is looking at. Riley mentioned loan portfolios and associated fees and information. Artificial intelligence can help ensure payments are made according to loan agreements, he said.

“It’s a way to complete the cycle, speed up processes and improve the end-to-end accuracy of the transaction record back to the legal contract,” Riley said.

Citigroup also plans to use large language models — an artificial intelligence algorithm that condenses vast sets of data — to digest laws and regulations in its countries of operation to ensure it complies with those rules. As a global bank, complying with each jurisdiction’s regulations can be onerous.

“If you’re a big bank and you want to make sure you’re meeting all the standards around the world, that’s very important,” Riley said. “This allows for a kind of map where you know you’re following laws and regulations.”

And while testing the technology, the bank makes sure it has the controls in place to fit Citigroup’s risk parameters. Now, Riley says, Citigroup employees have presented the bank with more than 350 AI use cases.