OpenAI is taking aim at some of Wall Street’s most labor-intensive tasks with a new version of ChatGPT designed to research companies, analyze financial data, and generate presentations that investment bankers rely on.
The product announced Thursday, called ChatGPT for Financial Services, is a customized version of the company’s enterprise product, ChatGPT Work, in collaboration with “design partners.” morgan stanley and evercoresaid Nick Turley, vice president of products at OpenAI. We use the artificial intelligence company’s latest and most advanced model, GPT-6 Astra.
The rollout moves OpenAI deeper into territory traditionally occupied by entry-level Wall Street bankers and the new graduates and associates known as analysts that the industry has hired for decades to research deals and write pitches. It also shows the company’s continued focus on enterprise products as it prepares for a blockbuster initial public offering.
“We’re effectively teaching ChatGPT to research like an analyst and back up its conclusions like an analyst,” Turley said at a briefing announcing the new product.
OpenAI has spent much of the last year competing for business customers in the highly competitive enterprise market, where it is trying to fend off rivals like Anthropic and Anthropic. google. Last year, Anthropic introduced Claude for Financial Services, a solution uniquely tailored to Wall Street.
OpenAI’s head of finance Sarah Friar told investors in August that the company’s enterprise business accounts for more revenue than its consumer business, which took off after the launch of ChatGPT in 2022.
Turley told reporters during a briefing that OpenAI plans to release customized solutions for “many sectors” beyond financial services.
In a live demonstration of the new product, Turley demonstrated a platform for analyzing potential M&A targets, pulling financial numbers from industry-standard data sources and creating pre-formatted PowerPoint decks based on the bank’s pre-formatted style guide.
“It’s very easy to create slides that look good, but much harder to create slides that actually make sense,” Turley said. “To get here, ChatGPT needed to select relevant peers. Prices needed to be pulled into a spreadsheet. Data had to be matched to charts and the declines and rebounds had to be accounted for.”
Bunker confusion?
The difference between this version and the product it is based on, ChatGPT Work, is the native data access from LSEG, Daloopa, and PitchBook, which not only provides financial statements, revenue records, etc. to the system, but also provides automatic access to users’ existing data subscriptions.
Other features built specifically for finance include citations, which allow users to trace data back to source files and audit charts, and administrative controls for sensitive transaction materials.
Mr. Turley said there was “massive demand” for this version of ChatGPT, initially aimed at investment banking and equity research, but declined to name the banks that signed on to the version.
When asked by CNBC if this latest version of ChatGPT will reduce the need for investment banks to hire junior bankers, Turley described the release as an efficiency boost that maximizes productivity per employee.
“If you look at the lives of analysts and bankers, depending on the industry, they work 100 hours a week,” Turley said. “Just as Microsoft Excel transformed the industry and enabled us to create better analyzes faster, these technologies will allow us to do the same.”
Still, the product raises fundamental questions about an industry long built on strict apprenticeships. If generative AI can perform multi-step tasks like research and pitchbook formatting in minutes, Wall Street will have to rethink how it trains the next generation of dealmakers and how many are needed.
Last month, Chris Churchman goldman sachs A partner in charge of one of the bank’s flagship AI projects has warned that automating tasks that help train junior bankers risks causing “cognitive atrophy” in the next generation of lenders.
“Inference remains important,” Churchman said at the time. “You have to reason[the problem]and frame it into an argument, and now you’re delegating the reasoning.”
