OpenAI is not quietly testing the waters in finance. The company has announced ChatGPT for Financial Services, a purpose-built product for investment banking and equity research teams that runs on GPT-6 Astra and ships with pre-integrated data from Daloopa, PitchBook, LSEG News, and Crunchbase. This is not a generic enterprise ChatGPT subscription with a finance-flavored skin. It is a targeted product built in direct collaboration with Morgan Stanley and Evercore, and the difference shows in the specifics.
What it actually does
The core pitch is straightforward: financial teams waste too much time wrestling with data access, MCP connectors, and source verification. ChatGPT for Financial Services tries to eliminate those friction points by hosting and indexing premium datasets directly on OpenAI infrastructure. That means no separate contracts with data providers, no connector setup, and lower latency on retrieval. Granular citations let analysts trace figures and claims back to specific tables and passages, which matters enormously in a business where a misattributed EBITDA adjustment can derail a deal.
GPT-6 Astra drives the reasoning layer. OpenAI claims it scores 69.9% on OfficeQA Pro, a benchmark that tests AI on complex financial tables, charts, and footnotes in documents like U.S. Treasury Bulletins. That compares to 60.2% for GPT-5.6 Sol. The model can also generate artifacts: documents, spreadsheets, and slides built from the analysis it produces. So instead of copying outputs into a deck manually, teams get finished materials directly. That is a meaningful workflow change for junior bankers spending hours on formatting.
The data stack
The built-in datasets cover earnings transcripts, financial statements, company fundamentals, and private company data. But OpenAI also knows firms already pay for their own subscriptions. So it is building shared sign-in and entitlement integrations with S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva, and Moody’s. Users can authenticate through their existing ChatGPT login and get automatic access to data they are already entitled to. That is a smarter approach than asking firms to re-license everything through a new vendor.
The broader connector ecosystem includes over 50 integrations, including Datasite, Box, Preqin, FactSet, and Intapp. OpenAI says it has also optimized the most commonly used financial MCP connections to reduce troubleshooting time. That is a low bar to celebrate, but anyone who has tried to get MCP connectors working reliably in a production environment knows it is a real problem.
Why Morgan Stanley and Evercore as design partners matter
Design partnerships in enterprise AI are often cosmetic. Not here. OpenAI says the collaboration with Morgan Stanley and Evercore directly shaped which workflows to prioritize and where the biggest pain points actually sit. Both firms have confirmed they are actively testing and influencing the product’s development. That gives OpenAI a credibility signal that competitors like Microsoft Copilot for Finance or Bloomberg’s AI initiatives will find difficult to match without similar institutional buy-in.
Who should care and who should wait
For large investment banks and equity research teams, this is worth a serious look. The combination of GPT-6 Astra’s reasoning, built-in data access, and enterprise governance controls addresses real operational problems. Smaller firms or those outside investment banking and equity research should probably wait. OpenAI is explicit that this launch focuses on those two categories first, with expansion into broader financial services to follow.
- Built-in data from Daloopa, PitchBook, LSEG News, and Crunchbase
- Entitlement integrations with S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva, and Moody’s
- 50+ connectors including FactSet, Preqin, Datasite, and Intapp
- GPT-6 Astra with 69.9% OfficeQA Pro benchmark score
- Granular citations linking outputs to source documents
- Enterprise security and governance controls with centralized access management
Pricing details are not public yet. Availability is through OpenAI’s financial services sales team, which suggests enterprise contracts rather than self-serve. That is standard for this market, but it also means adoption will be slow and sales-cycle dependent. Still, the product direction is clear. OpenAI is building domain-specific depth on top of its frontier models, and finance is the first serious test of whether that strategy actually converts into enterprise revenue.
