OpenAI Targets the Work Junior Bankers Do

OpenAI launched ChatGPT for Financial Services Thursday (Sept. 10), according to an announcement on its website. The product is a tailored version of ChatGPT Work, built with Morgan Stanley and Evercore as design partners. It runs on GPT-6 Astra, OpenAI’s newest model, and combines it with premium data from Daloopa, PitchBook, LSEG News and Crunchbase.

OpenAI hosts and indexes that data on its own infrastructure, rather than requiring banks to build separate connectors, per the announcement.

“We’re effectively teaching ChatGPT to research like an analyst and back up its conclusions like an analyst as well,” OpenAI Vice President of Product Nick Turley said, according to a Thursday CNBC report.

The tool targets LBO modeling, buyer screening, earnings analysis and pitchbook creation, with citations tracing each figure to its source filing, the OpenAI announcement said.

Anthropic built a similar product first. Claude for Financial Services connects to FactSet, S&P Capital IQ, Morningstar, PitchBook, Daloopa and Moody’s through pre-built MCP connectors, alongside 10 ready-to-run agent templates for pitchbooks, know your customer (KYC) screening and month-end close, according to a May 5 Anthropic announcement.

Citadel Head of Core Engineering Atte Lahtiranta, said in the announcement that analysts use Claude for Excel to “build and update coverage models, separate signal from noise, and pressure-test their work.”

The architectural split is real. Anthropic connects out to each provider’s own system. OpenAI licenses and hosts several datasets itself, giving it tighter control over retrieval and citations.

Startups Already Feel the Squeeze

OpenAI’s launch adds a second major lab to a fight AI native startups Rogo and Hebbia have faced since Claude for Financial Services shipped. Both startups turn written instructions into financial models, memos and pitch decks, plugging into a bank’s paid data and its own files. Both also run on Anthropic, OpenAI and Google models.

Rogo passed $50 million in annual recurring revenue, up more than threefold from $15 million at the end of 2025, and is on track for roughly $100 million by early next year, an unnamed source said, according to a Sept. 3 report from The Information.

Both companies said the labs can’t replace what they sell. Another unnamed source said in the same report by The Information that its agents behave “increasingly like your firm rather than like a generic Claude” once fed a bank’s own history, an argument both companies make in different words about why deep integration with a specific firm’s own data still matters more than a general-purpose model alone.

Fight Over Who Owns the Workflow

OpenAI’s choice to host some data directly, rather than connect out the way Anthropic does, signals a different bet on where control belongs. Selected datasets from Daloopa, PitchBook and LSEG News are bundled, indexed and hosted directly on OpenAI’s own infrastructure, requiring no separate contract or connector.

Other providers, including S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva and Moody’s, get a lighter sign-in integration instead, recognizing a user’s existing entitlements rather than hosting the data itself, according to OpenAI’s Thursday announcement.

For Rogo and Hebbia, the bigger threat may not be the labs at all. Their own customers are building internal AI agents on the same underlying models. According to The Information report, Farsight CEO Samir Dutta, who also competes with Rogo and Hebbia, said most banks still lack the scoped use cases needed for effective in-house tools.

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.