OpenAI Partners with Morgan Stanley to Launch Finance Edition of ChatGPT; GPT-6 Astra Targets Wall Street Junior Analysts

On September 10, 2026, OpenAI officially launched ChatGPT for Financial Services—a finance-industry customized version of its enterprise product ChatGPT Work. It uses GPT-6 Astra as the underlying model, with Morgan Stanley and boutique investment bank Evercore serving as design partners. The product has built-in financial data sources such as Daloopa, PitchBook, LSEG News, and Crunchbase; it can automatically generate editable financial models, research reports, and pitch materials according to each bank's own templates, and provides granular citation traceability. OpenAI explicitly describes the target scenario as taking on the core work of junior investment banking analysts.

How It Actually Works

The core mechanism of ChatGPT for Financial Services is to bind the reasoning capabilities of a large language model together with real-time access to professional financial data. The product connects to an external data ecosystem through more than 50 MCP connectors; partners include S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva, and Moody's, as well as FactSet, Preqin, Datasite, and other institutions—covering the main sources for daily data work in investment banking today. This means users do not need to switch between systems; data retrieval, analysis, and material generation can be completed within a single interface.

Nick Turley, OpenAI's vice president of product, demonstrated a complete workflow in the launch briefing: the system analyzes a potential M&A target, automatically pulls financial data from industry-standard data sources, and then generates formatted PowerPoint slides according to the bank's preset style guidelines. At the launch event, Turley said, "We are essentially teaching ChatGPT to research like an analyst and to support conclusions with data like an analyst." He also emphasized, "Making good-looking slides is easy, but making slides that actually make sense is much harder"—to complete this task, the system must autonomously select comparable peers, import price data into spreadsheets, check the consistency between charts and original data, and explain stock price movements.

On compliance and security, the product supports SAML single sign-on, SCIM permission management, role-based access control, encryption of data at rest and in transit, and audit log export; these features directly correspond to the regulatory compliance requirements of the financial industry. GPT-6 Astra scored 69.9% on OpenAI's OfficeQA Pro benchmark, higher than the previous generation GPT-5.6 Sol's 60.2%.

Gains and Losses for Stakeholders

For financial data providers, the impact was felt the same day. FactSet Research shares fell 5.3% to $264.32 on September 10; on the same day, the S&P 500 software and services index fell 1.5%, Salesforce dropped 4%, and ServiceNow and Intuit each fell about 5%. OpenAI built access to data such as PitchBook and LSEG News directly into the product, rather than directing users to purchase separate subscriptions. This shakes the core business model of financial data intermediaries—charging high subscription fees by using exclusive data access as leverage. When users can retrieve multi-source data through ChatGPT's unified interface, the bargaining power of a single data provider will narrow. OpenAI maintains partnerships with institutions such as FactSet and S&P Global at the same time.

For junior investment banking employees, Wall Street junior analysts and associates have traditionally handled data cleanup, comparable company analysis, financial model building, and pitch material production—exactly the types of tasks that ChatGPT for Financial Services demonstrated one by one in its launch demo. The value of junior positions lies not only in producing documents, but also in accumulating an understanding of industries, deals, and clients through repeatedly performing these tasks.

For OpenAI itself, this launch has a clear commercial motive. CFO Sarah Friar disclosed to investors in August this year that the company's enterprise business revenue had surpassed its consumer business. With OpenAI widely expected to complete an IPO soon, the financial services vertical edition is not just product expansion but also part of its enterprise revenue narrative: showing potential investors that the company can systematically monetize AI capabilities into high-value industry solutions, rather than relying only on monthly subscription revenue. Turley revealed that financial services is only the starting point, and OpenAI plans to launch similar customized versions for multiple industries.

For developers and system integrators, the open architecture of more than 50 MCP connectors provides interface opportunities, but it also means OpenAI has standardized the connection layer, and third-party differentiation space will narrow as the scope of OpenAI's direct connections to data sources expands.

Anthropic Moved First, but OpenAI Has Design Partner Backing

ChatGPT for Financial Services is not Wall Street's first major AI platform. Anthropic launched Claude for Financial Services as early as July 2025 and has already deployed it in production at institutions such as JPMorgan, Goldman Sachs, Citi, AIG, and Visa—meaning Anthropic has been about a year ahead in institutional penetration among major U.S. banks. In May 2026, Anthropic further released 10 ready-made AI agent templates for investment banking, asset management, and insurance workflows, and established data partnerships with Moody's, FactSet, Morningstar, and S&P Global, fully integrating with Microsoft 365. At the launch event, JPMorgan CEO Jamie Dimon demonstrated live how to build a Treasury asset swap analysis dashboard in 20 minutes.

There are several specific differences in the competitive landscape between the two. First, the depth of design differs: OpenAI's Morgan Stanley and Evercore are partners involved from the product design stage, meaning the product has built-in understanding of these two institutions' actual workflows; Anthropic's big-bank cooperation is more reflected in deployment-level implementation rather than co-design. Second, institutional coverage differs: Anthropic already has production deployment records at multiple top-tier big banks, while OpenAI's currently public list of institutional partnerships remains mainly design partners. Third, the model foundation differs: GPT-6 Astra is OpenAI's most advanced model, while Anthropic's corresponding financial scenarios mainly use Claude Opus 4.7.

What Is Most Likely to Happen Next

From an industry perspective, the real significance of ChatGPT for Financial Services lies in choosing an extremely persuasive proof scenario. Investment banking is a highly rule-based, document-intensive, and heavily regulated industry—vertical AI deployment is extremely difficult. If AI can be adopted at the institutional level in this scenario and pass compliance review, its migration to similar industries such as law, accounting, and consulting will greatly reduce decision friction. OpenAI's establishment of a benchmark in the financial industry has strategic value beyond finance itself.

Regulatory compliance is the most critical variable. The financial industry has regulatory-level requirements for model interpretability and conclusion traceability, far exceeding the voluntary adoption threshold of general enterprise SaaS. Granular citation traceability is a feature explicitly emphasized by ChatGPT for Financial Services, indicating that OpenAI has incorporated compliance into product design. How the U.S. SEC and financial regulators in various jurisdictions determine the compliance of AI-generated materials will determine how deeply the product can be adopted in core business scenarios.

The pricing model is another signal. Bloomberg Terminal's per-seat annual fee has historically exceeded $20,000; financial data subscriptions are a high-barrier revenue source. If OpenAI prices by bundling data access, it will compete head-on with existing data providers; if it maintains a usage-based API model, it will look more like adding a reasoning layer on top of data. The current market reaction of FactSet and other stocks leans toward the former judgment, but the final pricing strategy will be determined by commercial negotiations.

For financial institutions currently evaluating AI tools, the most pragmatic execution judgment is: do not view this launch as a multiple-choice question of whether to adopt AI, but understand it as an implementation question of at which workflow node to introduce it, whose solution to choose, and how much review cost to bear. Whether with OpenAI's or Anthropic's solution, junior analyst positions will not immediately disappear because of a single launch, but within the same amount of time, there will be a structural gap in output efficiency between analysts who can use these tools and those who cannot. What institutions truly need to build is a judgment mechanism for reviewing the quality of AI-generated materials—precisely the part neither product currently has built in and that institutions must construct themselves.