OpenAI has launched ChatGPT for Financial Services, a specialised ChatGPT Work experience aimed at investment banking and equity research teams. The product combines GPT-6 Astra with financial data, connected sources and the security and governance controls that financial institutions expect when working with sensitive information.
The launch is notable because it packages model capability and licensed information into a tailored workflow. OpenAI says the product can help teams develop research, financial models and customised client materials. Its early design work with Morgan Stanley and Evercore focused on the practical obstacles in financial work: access to dependable data and producing deliverables that meet a firm's standards.
Premium data inside the workflow
ChatGPT for Financial Services includes built-in data from Daloopa, PitchBook, LSEG News and Crunchbase. OpenAI says the data covers earnings transcripts, financial statements, company fundamentals and private companies. Rather than requiring teams to negotiate separate connectors before getting started, the company says it indexes and hosts that material to improve retrieval, latency and citation behaviour.
The product also supports existing subscriptions. OpenAI is working on entitlement integrations with S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva and Moody's, allowing providers to recognise a user through ChatGPT sign-in where the user already has access. For financial teams, this distinction matters: a useful assistant needs to respect both the contracts that govern data and the evidence trail needed to assess a conclusion.
OpenAI says it has also improved reliability for commonly used financial-services MCP connections, including S&P Global and FactSet. Its broader connector ecosystem includes Datasite, Box, Preqin and Intapp. The central promise is not simply a greater number of integrations, but a more direct route from an entitlement to an analysis that can be checked.
Analysis with evidence and artefacts
OpenAI describes GPT-6 Astra as the reasoning layer behind information retrieval, financial reasoning and document creation. The system is intended to navigate figures, tables and supporting notes, then turn the work into documents, spreadsheets and slides. Users can trace figures and claims to tables and passages, with supporting material highlighted for review.
That emphasis on verification is crucial in finance. A generated chart or valuation model is useful only when a reviewer can inspect the inputs and understand the assumptions. OpenAI says users can explore underlying data and sources, compare company performance in interactive charts and create client-ready outputs from the analysis.
The tailored experience also includes administration for templates. Firms can publish Excel, Word and PowerPoint templates and style guides through a dedicated page, then have teams use those materials when producing valuation models, research notes and pitchbooks. This may help institutions standardise output while still letting individual teams use natural-language workflows.
Governance remains central
OpenAI says the product builds on ChatGPT Enterprise controls including SAML SSO, SCIM provisioning and role-based access controls. Business data is not used to train models by default, is encrypted in transit and at rest, and workspace retention can be configured by administrators. Compliance teams can export supported workspace logs through the OpenAI Compliance Platform for their own investigation and audit processes.
These controls do not eliminate the need for a firm's own review. Institutions considering the product will need to determine where it fits in analyst supervision, data-entitlement management and record-retention policies. They will also need to test its outputs against the sources and calculations that underpin their existing process.
OpenAI's announcement positions ChatGPT for Financial Services as a purpose-built environment rather than a generic chatbot with a finance prompt. If its data integrations, citations and template workflows hold up in day-to-day use, it could give financial teams a faster way to move from research to a reviewable client deliverable without separating the analysis from the evidence behind it.
What adoption will require
Financial institutions will need a structured evaluation that covers data entitlements, model output, user permissions and supervision. A pilot should include common research and modelling tasks, with experienced reviewers checking citations, calculations and the treatment of ambiguous source material. This establishes where the product saves time and where existing controls must remain unchanged.
The opportunity is to combine speed with an audit-ready process. If teams can ask a question, inspect supporting evidence and turn a validated analysis into a standardised deliverable, the product may reduce repetitive work without loosening the governance that financial analysis demands.
Leaders should define ownership for data-provider relationships, validate template controls and establish a review process for client-facing output. Those operational choices will determine whether the new specialised environment can fit safely into the firm's established research, compliance and approval model.