Citi's CTO says these are the kinds of jobs he's eliminating with AI

Citi is well within the throes of its AI transformation. More than 180,000 Citi employees use the bank's AI tools, the majority of which do so on a regular basis. Speaking at a symposium by AI insights firm Evident released on YouTube this week, Citi CTO David Griffiths revealed some of the most promising use cases the bank has found for its AI tools, as well as the team that's driving them.

Click here to join the bubble by eFinancialCareers, our new anonymous community. ✍️

Griffiths said Citi has a "compendium" of AI use cases that is being continually added to by a group of 4,200 "AI accelerators." These employees spread awareness of AI use cases for the bank's employees in various ways, including creating "short bite-size videos" of themselves doing "cool stuff" with AI. It's not the only AI accelerator in banking; JPMorgan has one which deploys itself into different teams in the bank to help them increase their AI usage.

The most popular tool in Citi's compendium does the work of junior bankers. Griffiths said there's a tool that pulls data from various sources and distills it into a PowerPoint that it generates for you. It's incredibly popular: "I'm always suprised by how everyone says 'that's the one I want,'" he said. Griffiths suggested that the PowerPoints generated do still need human contribution, however; he said that "there's still a last mile where you have to go in and do it yourself," and that "if anyone can help us make prettier PowerPoints, that would be amazing."

What other kinds of things does Citi's compendium cover? Griffiths said Citi is exploring AI applications in teams with "large populations," the most promising of which is its technology function because "the AI models are so incredible at coding." This appears to be the consensus; MDs at Bank of America and Morgan Stanley each lauded the powers of AI coding at this month's Quant Strats Europe conference; Stephan Kessler, global head of quant research for the latter bank's quantitative investment strategies team called CoPilot a "revelation" for coding in Python. Griffiths said that the likes of Claude and ChatGPT are "highly optimized for coding output, and they're getting better all the time." 

It's not just the code itself Citi is writing with AI. Griffiths said the bank has "built a bunch of things" including AI "code review documentation," which has helped Citi's platform complete over one million code reviews. 

Citi is looking to distil tasks that took days into a matter of minutes. For example, he said crafting a memo used to take "a couple of days," including half a day to actually write it; using AI, the writing and editing takes ~20 minutes. Citi's AI platform also has a 'middle layer' that performs multiple tasks including "authentication, entitlement, logging telemetry [and] monitoring costs." The implication is that middle office jobs are also susceptible to the AI axe. 

If those jobs are susceptible, then which jobs are safe? Marco Argenti, chief information officer at Goldman Sachs, also spoke at the symposium and gave an answer a little at odds with Griffiths; he said there's an inside joke at Goldman Sachs that goes "would the last person walking out of Goldman Sachs be an engineer or a trader?" Argenti agreed that operations staff are susceptible to AI in the short term, mentioning the likes of KYC onboarding, customer service and IT support as potential risky career paths.

Have a confidential story, tip, or comment you’d like to share? Contact: WhatsApp: http://wa.me/442079977910 (+44 20 7997 7910), Telegram: @AlexMcMurray, Signal: @AlexMcMurrayEFC.88 Click here to fill in our anonymous form, or email editortips@efinancialcareers.com.

Bear with us if you leave a comment at the bottom of this article: comments are moderated intermittently by human beings. Sometimes these humans might be asleep, or away from their desks, so it may take a while for your comment to appear. You must take sole responsibility for comments you post on this site. We will take reasonable steps to weed out anything that we consider to be offensive or inappropriate.