OpenAI Targets the Work Junior Bankers Do

OpenAI unveiled ChatGPT for Financial Services on September 10, reported CNBC. The product, customized with input from Morgan Stanley and Evercore, leverages OpenAI's latest model, GPT-6 Astra, and integrates premium data from Daloopa, PitchBook, LSEG News, and Crunchbase. OpenAI hosts and indexes this data internally, eliminating the need for banks to develop separate connectors.

OpenAI's Vice President of Product, Nick Turley, explained that the tool aims to replicate an analyst's research and analysis capabilities. Analysts can utilize the tool for tasks like building LBO models, evaluating buyers, analyzing earnings, and creating pitchbooks, with citations referencing the source documents. Anthropic had previously developed a comparable product, Claude for Financial Services, which connects to fact-based databases such as FactSet, S&P Capital IQ, and Morningstar through pre-built integrations, alongside ready-to-use templates for pitchbooks, KYC screening, and month-end close.

According to Citadel's Head of Core Engineering, analysts use Claude to refine Excel models by filtering out irrelevant information and validating their work. However, OpenAI's approach differs significantly from Anthropic's. While Anthropic connects to each provider's system separately, OpenAI licenses and hosts several datasets internally, granting them greater control over data retrieval and citation tracking.

This architectural distinction places OpenAI in a more advantageous position, as it can offer tighter control over data and citations. The launch of ChatGPT for Financial Services intensifies competition among AI-native startups like Rogo and Hebbia, who cater to the banking sector's demand for AI-driven financial modeling, memo writing, and pitch deck creation.

Both startups rely on Anthropic and Google models to process written instructions into professional financial documents. They have already achieved significant traction, with Rogo's annual recurring revenue reaching $50 million, a threefold increase from $15 million in 2025, and on track to hit $100 million by early 2026. Despite the competition, both companies emphasize that their solutions cannot replace the comprehensive services they provide.

Furthermore, both companies assert that deep integration with a specific bank's data is crucial, as it enables the AI agents to behave more like the bank and offer tailored solutions rather than relying on a generic model. OpenAI's decision to host certain datasets directly, instead of merely connecting to external providers like Anthropic does, reflects a strategic shift in control.

By bundling and indexing selected data from Daloopa, PitchBook, and LSEG News on its own infrastructure, OpenAI reduces the need for separate contracts and connectors. Still, OpenAI relies on lighter sign-ins with other providers, such as S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva, and Moody's, which only recognize the user's existing data entitlements.

This approach may be a strategic gamble by OpenAI, as many banks currently lack the specific use cases required for effective in-house AI tools. Farsight CEO Samir Dutta told The Information that many financial institutions still need to develop the precise use cases necessary for successful internal AI systems.