Teaching ChatGPT to Research Like an Analyst: OpenAI Ships ChatGPT for Financial Services

On September 10, 2026, OpenAI pulled back the curtain on ChatGPT for Financial Services, a purpose-built version of its enterprise workspace that aims squarely at the most labor-intensive work on Wall Street: researching companies, building financial models, and generating the pitchbooks that investment banks have depended on armies of junior bankers to produce. The launch is OpenAI’s most aggressive vertical play yet, arriving just days after GPT-6 Astra reached general availability and as the company gears up for what is widely expected to be a blockbuster IPO.

What shipped

At its core, the product is a tailored ChatGPT Work experience powered by GPT-6 Astra, OpenAI’s newest frontier model. What separates it from the horizontal ChatGPT Work offering is native financial data access. Datasets from Daloopa, PitchBook, and LSEG News — covering earnings call transcripts, financial statements, company fundamentals, and private-market data — are indexed and hosted on OpenAI’s own infrastructure. Teams can start querying them immediately, with no separate contracts to negotiate and no connectors to wire up.

For firms that already hold premium subscriptions, OpenAI is building shared sign-in and entitlement integrations with S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva, and Moody’s. Under that model, providers recognize users through their ChatGPT sign-in and automatically unlock data those users are already entitled to. OpenAI has also optimized widely used MCP connectors like S&P Global and FactSet for the product, with a broader ecosystem of more than 50 connectors including Datasite, Box, Preqin, and Intapp.

The product was shaped by a design partnership with Morgan Stanley and Evercore, whose early feedback steered the starting point toward investment banking and equity research — the segments where reliable data access and high-quality artifact creation hurt the most. OpenAI says that partner work will feed directly into post-training, product improvements, and expansion into other financial services categories.

The benchmark OpenAI wants bankers to see

OpenAI positions GPT-6 Astra as state-of-the-art across the three capabilities it considers core to finance work: information retrieval, financial reasoning, and artifact generation. The company’s headline number comes from OfficeQA Pro, an internal benchmark testing whether AI agents can find and analyze information across U.S. Treasury Bulletins — including complex financial tables, charts, and supporting footnotes. There, GPT-6 Astra scores 69.9%, versus 60.2% for its predecessor GPT-5.6 Sol.

The retrieval story matters more than the raw score. Because the data is indexed on OpenAI’s infrastructure, the system can issue granular citations — a banker running a P&L normalization analysis can inspect the reconciliation and notes behind an adjusted EBITDA figure, see exactly which costs were excluded, and decide how to use it in a valuation. In an industry where every number in a deck has to survive an audit, traceability is the difference between a demo and a tool.

Administrators can publish Excel, Word, and PowerPoint templates through a dedicated admin page. With firm templates and style guides configured, teams can convert analysis into valuation models, research notes, and pitchbooks in the bank’s own format — the artifact-generation piece that Turley demonstrated live, showing the platform analyze a potential M&A target, pull financials from industry-standard sources, and emit a formatted PowerPoint built on the bank’s preformatted style guide.

“Research like an analyst”

“We’re effectively teaching ChatGPT to research like an analyst and back up its conclusions like an analyst as well,” Nick Turley, OpenAI’s vice president of product, said during the launch briefing. His demo emphasized that the hard part isn’t pretty slides. “To get here, ChatGPT had to choose the relevant peers. It had to pull the prices into a spreadsheet. It had to check the chart against the data, and it had to explain the sell-off and the rebound.”

Asked directly by CNBC whether the product would reduce the need to hire junior bankers, Turley framed it as an efficiency gain: “If you study the life of an analyst or of a banker, depending on the industry, they’re working 100-hour weeks. I think in the same way that Microsoft Excel transformed the industry and allowed them to produce better analysis faster, you will see technology like this do the same.”

The apprenticeship problem

The uncomfortable subtext is what happens to Wall Street’s training pipeline. The industry has run on an apprenticeship model for decades: analysts learn the business by doing the grunt work that this product now automates. Last month, Chris Churchman, the Goldman Sachs partner in charge of one of the bank’s flagship AI projects, warned that automating the tasks which train junior bankers risks causing “cognitive atrophy” in the next generation of financiers. “Reasoning is still important,” he said. “You still need to reason about problems and structure it into an argument, and now we’re delegating reasoning.”

There is also a competitive dimension. Anthropic shipped its own tailored Claude for Financial Services last year, and enterprise is now the main battleground — OpenAI CFO Sarah Friar told investors in August that the enterprise business already accounts for more revenue than the consumer side. Turley told reporters that tailored solutions for “a number of sectors” beyond finance are coming.

Security and compliance

Because this is banking, the compliance scaffolding is substantial. The product inherits ChatGPT Enterprise’s SAML SSO, SCIM provisioning, and role-based access controls. Business data is not used to train OpenAI models by default and is encrypted at rest and in transit. Administrators can configure workspace retention, export supported workspace logs through the OpenAI Compliance Platform into existing audit workflows, manage access to skills and apps by role, enable or disable app read/write actions, and spin up multiple workspaces to enforce information barriers between deal teams — a hard requirement for any bank running competing mandates.

Availability is by eligibility: OpenAI says the product is open to qualifying financial institutions, directing interested firms to contact the company or their account team. No pricing was disclosed publicly.

Why it matters

This launch compresses three trends into one announcement. First, verticalization: the generic chatbot era is giving way to domain-specific products where the moat is data plumbing — entitlements, connectors, citations — rather than raw model quality. Second, the enterprise revenue race: with Anthropic and Google fighting for the same contracts, finance is the highest-value beachhead, and OpenAI just landed with both feet. Third, the labor question: if a tool can do in minutes what took an analyst a week, the industry’s hundred-hour-week apprenticeship model stops being a rite of passage and starts being a design choice. Turley’s Excel comparison is apt in ways he may not have intended — Excel didn’t eliminate analysts, but it permanently changed what they do all day. This will too.