ChatGPT for Financial Services: From Cited Data to Models and Pitchbooks

The financial-services problem it targets

Financial analysis loses substantial time to locating sources, reconciling which number is current, moving data into a model and repackaging the result as research or a pitchbook. OpenAI designed ChatGPT for Financial Services to combine data, reasoning and artifact creation in one ChatGPT Work experience.

The first focus is investment banking and equity research, shaped with Morgan Stanley and Evercore. The purpose is not autonomous investment decisions; it is reducing retrieval and formatting work so experts can spend more time on judgment and debate.

Built-in data versus another connector

OpenAI says selected datasets are indexed and hosted on its infrastructure, allowing eligible teams to start without negotiating a separate contract or building a connector for every source. The launch names providers such as Daloopa, PitchBook and LSEG News, while the Help Center lists additional sources and access constraints.

Included coverage may differ from a provider's full product. Some sources require an existing subscription and permissions; some datasets have delays or quotas; availability can change.

The workflow: research, model, challenge, present

• Research filings, transcripts, fundamentals and news with traceable sources
• Build or update valuation, scenario, normalization, peer-comparison or LBO models
• Ask for assumptions, sensitivities, downside cases and evidence against the thesis
• Turn analysis into research notes, client memos and pitchbooks using firm templates
• Review formulas, citations, dates, units, currencies and language before delivery

Why templates and granular citations matter

Administrators can publish Excel, Word and PowerPoint templates and firm style guides so teams produce material in an approved format. Granular citations can point back to supporting tables or passages.

A citation is not a correctness guarantee. Analysts still need to check source, reconciliation, notes, definitions, periods and units—and distinguish actuals, estimates and AI-derived assumptions.

Governance for financial and client data

• Enforce role and provider-entitlement permissions
• Separate MNPI, client-confidential, internal-research and externally shareable data
• Apply SAML SSO, SCIM, RBAC and workspace retention policies
• Export supported logs to the Compliance Platform for audit and investigation
• Require accountable review before any client material or investment conclusion
• Record source date, model, template, formula changes and approver

Availability and important limits

• Access is for eligible financial institutions through OpenAI Sales or an account team
• It applies to the full workspace; the Help Center says standard Enterprise and Financial Services seats cannot be mixed in one workspace
• Some data sources are limited by country, profession, subscription or permissions
• The product supports research and is not financial or investment advice
• International institutions should confirm local-market and language coverage before rollout

Pilot without putting money or clients at risk

Start with a reviewable use case such as an earnings summary or peer table. Use historical or public data, build a gold set and define citation and formula checks before introducing client information or MNPI. Measure turnaround time, error rate, reviewer time and adoption against the current process.

Our ChatGPT Work × Claude Cowork workshop helps teams design research workflows, templates, skills, connectors and approval gates around real work rather than isolated demos.