Bankers and equity analysts got a version of ChatGPT built for their work on September 10, packaged with licensed market data that OpenAI hosts itself.
The design work started with two customers. Teams at Morgan Stanley and Evercore described a day lost to two things: data they could not reach reliably, and documents that took too long to produce. Investment banking and equity research became the first scope as a result.
Three providers feed the product: Daloopa, PitchBook and LSEG News. The corpus takes in transcripts of earnings calls, financial statements, fundamentals data and private-company records. OpenAI indexes and hosts all of it, which the company says sharpens retrieval and citations and spares customers the usual contract talks and connector builds. Someone normalizing a P&L can open the notes behind an adjusted EBITDA line and follow each number back to its table.
The assistant runs on GPT-6 Astra natively. Firms can publish their own Excel, Word and PowerPoint templates through an admin page, so output arrives in house style as valuation models, research notes and pitchbooks.
Firms that already license their own data will be able to keep using it. For that piece, five providers have joined: S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva and Moody’s. The plan is for each to recognize a ChatGPT sign-in and hand over whatever entitlements the customer already holds. More than 50 connectors are already shipping, including Datasite, Box, Preqin and Intapp.
Security carries over from ChatGPT Enterprise. Sign-on runs on SAML and provisioning on SCIM, with roles limiting what each person can open. Traffic and stored data are encrypted, retention windows are admin-set, and compliance teams can pull logs into their own tooling. Client business data stays out of training by default, and firms can wall off teams in separate workspaces.
