Invest Independently confirmed 4 publishers 3 min read Published Sep 10, 2026, 7:29 PM PDT

OpenAI brings ChatGPT to Wall Street 14 months after Anthropic's Claude for financial analysis

OpenAI's finance product was designed with Morgan Stanley and Evercore and is sold only to institutions the company clears in advance. Nick Turley said demand was heavy and would not name a bank that has bought it.

The Investor · Invest desk

Send a correction

What happened

OpenAI unveiled ChatGPT for Financial Services on Thursday, a tailored version of its ChatGPT Work enterprise product running GPT-6 Astra and built with design partners Morgan Stanley and Evercore.
Access requires a ChatGPT enterprise account and is open only to eligible institutions, which have to speak directly with OpenAI to get cleared.
Banks connect their own Bloomberg or FactSet subscriptions alongside pre-loaded Daloopa, PitchBook, Crunchbase and LSEG News feeds, with around 50 connectors in total through MCP.
Turley said there was a ton of demand for the product, which is aimed first at investment banking and equity research, and declined to name any bank that has signed on.

Why it matters

What it limits constraint Because every buyer has to be cleared in a direct conversation, the pace of adoption is set by how fast OpenAI can process institutions.
What it costs cost A bank cannot know what a deck costs until an analyst chooses an effort level, which pushes part of the bill from a fixed per-seat licence into a usage line that finance has to forecast.
Who has to decide decision Heads of banking now have to size next year's analyst class for the work that remains. The training that work provided is what Lau argues has been automated.
What doesn't add up contradiction Turley's Excel comparison says output per banker rises while headcount holds; DeStefano and Lau say efficiency improves but judgement stays human. The sources do not agree on whether hiring falls.

Next story · Build · yesterday

Codex agent solves 49 of 50 unpublished physics problems for under $3 each

Confirmed2 publishers

A user picks high, medium or low, the higher setting spends more tokens and costs more, and OpenAI says the higher setting can produce a better output ^[20]. The marginal cost of a deck therefore sits with whoever is building it at two in the morning. Those sample slides took Astra about 10 minutes ^[9]. OpenAI did not say at which setting.

The data bill stays where it was. A bank plugs in the market data subscriptions it already pays for ^[5], so what OpenAI adds to the invoice is enterprise seats plus token consumption that OpenAI does not set.

Anthropic debuted a version of Claude for financial analysis in July 2025 ^[4]. CNBC's report on the OpenAI launch is dated 10 September 2026 ^[6], about 14 months later ^[19]. Finance is one of three verticals OpenAI picked, with cybersecurity and software engineering ^[3], and Turley said tailored products for "a number of sectors" will follow ^[7].

"This is the canonical product we are hoping the industry adopts," Turley said ^[16], and OpenAI wants it to be "the one product" a bank with tens of thousands of employees ever needs ^[17]. Set against that ambition, the disclosure is thin: no price, no seat count, and no customer beyond the two design partners, since Turley said there was "a ton of demand" and then declined to name a bank that had signed ^[10]. Sarah Friar told investors in August that OpenAI's enterprise business accounted for more revenue than its consumer business ^[21]. That figure covers every enterprise product the company sells.

Asked whether banks would need fewer junior hires, Turley went to productivity. "If you study the life of an analyst or of a banker, depending on the industry, they're working 100-hour weeks," he said ^[11], and "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" ^[12]. A 100-hour week is more than 14 hours a day, seven days a week ^[24].

Theodora Lau, founder of Unconventional Ventures, told American Banker that "We are removing the ability for the junior bankers to learn what a wrong model looks like" ^[13] and that "Ten years down the road, you'll end up with managing directors who have never built a model from scratch. That's what apprenticeship is supposed to be for, and it just got automated by OpenAI" ^[14]. Annie DeStefano, a fintech and banking consultant who previously worked at Goldman Sachs and Silicon Valley Bank, told the same publication she does not see ChatGPT replacing junior analyst work entirely ^[23].

My read is that the clearance process sets how fast this sells: every buyer arrives through a direct conversation with OpenAI ^[8], and American Banker said the company did not immediately respond when asked what determines eligibility ^[18]. DeStefano's counter-argument is that the gate is the product. She said the depth regulated industries require for AI adoption will absolutely require this type of industry-led design partnership ^[22]. The reading fails if a bank with tens of thousands of employees turns up next quarter with thousands of seats live and a token bill to match.

What to watch

A named bank disclosing a seat count or a token spend line for ChatGPT for Financial Services.
Whether OpenAI publishes the criteria that make an institution eligible after declining American Banker's question.
Which of the two remaining verticals, cybersecurity or software engineering, ships next, and whether Anthropic updates Claude for Financial Services in response.

Information, not investment advice. This story is general information about companies and markets, drawn from cited reporting. It is not a recommendation to buy, sell or hold any security; check the cited sources before acting.

Clarity's read

What the record supports and how the coverage leans. The claims behind it follow.

Reality

Evidence55
Adoption18
Hype gap+34
Incentives80
Confidence64

Perspective Coverage

4 publishers

Builder: Builder 31%
Operator: Operator 39%
Investor: Investor 30%

Why these scores

Claim ledger

Ranked by verification strength, evidence, and original report placement.

[1]
OpenAI unveiled a new version of ChatGPT geared toward big banks and other financial firms, powered by its new GPT-6 Astra model.
Reported SupportedView cited source
[2]
The product, unveiled Thursday and called ChatGPT for Financial Services, is a tailored version of OpenAI's enterprise product ChatGPT Work, made with "design partners" Morgan Stanley and Evercore according to OpenAI Vice President of Product Nick Turley, and it uses GPT-6 Astra.
Reported SupportedView cited source
[3]
Turley said finance is one of the industry verticals OpenAI has chosen to build specific products for, along with cybersecurity and software engineering.
Reported SupportedView cited source
[4]
Anthropic already offers a version of Claude for financial analysis, which it debuted in July 2025.
Reported SupportedView cited source
[5]
Banks can connect their existing subscriptions to datasets from companies such as Bloomberg or FactSet; ChatGPT also offers pre-loaded data from Daloopa, PitchBook, Crunchbase and LSEG News, plus around 50 total connectors through MCP, and provides detailed citations when it uses that data.
Reported SupportedView cited source
[6]
CNBC published its report on ChatGPT for Financial Services on 10 September 2026.
Reported SupportedView cited source
[7]
Turley told reporters during the briefing that OpenAI plans to release tailored solutions for "a number of sectors" beyond financial services.
Reported SupportedView cited source
[8]
Banks need a ChatGPT enterprise account to access the new experience, and it is only available for "eligible institutions" that must speak directly with OpenAI to get cleared for access.
Reported SupportedView cited source
[9]
OpenAI showed reporters a few sample presentation slides that it said took about 10 minutes for Astra to create.
Reported SupportedView cited source
[10]
Turley said there was "a ton of demand" for this version of ChatGPT, which is initially geared toward investment banking and equity research, but he declined to name banks that have signed on for it.
Reported SupportedView cited source
[11]
Turley said: "If you study the life of an analyst or of a banker, depending on the industry, they're working 100-hour weeks."
Reported SupportedView cited source
[12]
Turley said: "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."
Reported SupportedView cited source
[13]
Theodora Lau, founder of Unconventional Ventures, told American Banker: "We are removing the ability for the junior bankers to learn what a wrong model looks like."
Reported SupportedView cited source
[14]
Lau told American Banker: "Ten years down the road, you'll end up with managing directors who have never built a model from scratch. That's what apprenticeship is supposed to be for, and it just got automated by OpenAI."
Reported SupportedView cited source
[15]
CNBC reported that the rollout puts OpenAI deeper into territory traditionally occupied by Wall Street's entry-level bankers, the analysts and associates the industry has employed for decades to research deals and create pitchbooks.
Reported SupportedView cited source
[16]
Turley said: "This is the canonical product we are hoping the industry adopts."
Reported SupportedView cited source
[17]
OpenAI hopes its offering will be "the one product" large banks, or those with "tens of thousands of employees," will ever need.
Reported SupportedView cited source
[18]
American Banker reported that OpenAI did not immediately respond to a request for comment on what determines eligibility for the product.
Reported SupportedView cited source
[19]
OpenAI's finance product arrived about 14 months after Anthropic's July 2025 debut of Claude for financial analysis.
Derived Supported
[20]
With a toggle, users can specify whether the Astra model should approach a task with high, medium or low "effort"; the higher the effort, the more tokens used and the more it costs, and OpenAI says higher effort could mean a better quality output.
Reported SupportedView cited source
[21]
Sarah Friar, OpenAI's finance chief, told investors in August that the company's enterprise business accounted for more revenue than its consumer business.
Reported SupportedView cited source
[22]
Annie DeStefano, a fintech and banking consultant with prior experience at Goldman Sachs and Silicon Valley Bank, said: "The depth [that] regulated industries require for AI adoption will absolutely require this type of industry-led design partnership."
Reported SupportedView cited source
[23]
DeStefano said she does not see ChatGPT replacing junior analyst work entirely, but sees potential for increases in efficiency.
Reported SupportedView cited source
[24]
A 100-hour week is more than 14 hours a day across seven days.
Derived Supported

Sources

4 independent publishers whose own reporting we read for this story.

americanbanker.com
1 article · September 10, 2026
OpenAI launches financial tool for Wall Street bankers
cnbc.com
1 article · September 10, 2026
OpenAI targets work of Wall Street junior bankers with new ChatGPT for Financial Services
fortune.com
2 articles · September 10, 2026
OpenAI courts Wall Street with ChatGPT for financial services, developed with Morgan Stanley
pymnts.com
1 article · September 10, 2026
OpenAI Focuses On Investment Banking Ahead of Wider Financial Services Rollout

Clarity analysis

How the coverage leans and what the record actually supports. Read the methodology

Perspective Coverage

How much of the coverage speaks to builders, operators and investors. It measures emphasis, not whether the story is right.

Builder31%

Fortune is the only outlet in our coverage that describes the machinery. Banks attach their own Bloomberg or FactSet subscriptions; Daloopa, PitchBook, Crunchbase and LSEG News come pre-loaded; there are around 50 connectors through MCP, and output lands as PowerPoint, Excel or a small web dashboard. The effort switch is high, medium or low, where higher effort spends more tokens and costs more, and OpenAI says only that it could improve quality. CNBC's account adds the demo mechanics -- Turley said 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." Nobody outside OpenAI has run it.

Operator39%

Getting this is not a click. You need a ChatGPT enterprise account, and the product goes only to "eligible institutions" cleared in a direct conversation with OpenAI. American Banker asked what makes an institution eligible and had no answer by publication. Fortune lists what an administrator can set: SAML SSO, SCIM provisioning, role-based access, configurable retention, and control over which roles reach which skills and apps. The training question is live in the same reporting. Theodora Lau, founder of Unconventional Ventures, told American Banker: "We are removing the ability for the junior bankers to learn what a wrong model looks like." Annie DeStefano, who worked at Goldman Sachs and Silicon Valley Bank, does not expect junior analyst work to disappear and says sensitive transactions will still need "continued judgement and discernment by trained professionals, within an established risk framework of an institution."

Investor30%

CNBC carries the money context: Sarah Friar told investors in August that OpenAI's enterprise business now brings in more revenue than consumer, and the network places this launch in the run-up to an expected IPO. Against that, the demand evidence is a quote. Turley said there was "a ton of demand" and would not name a bank that has signed on, and the two firms OpenAI does name, Morgan Stanley and Evercore, are described as design partners rather than customers. Fortune states the ambition plainly: OpenAI wants to be "the one product" banks with tens of thousands of employees ever need, in a category Anthropic entered in July 2025.

Reality

Evidence, adoption, hype gap, incentives and confidence, each read independently from the current record on the scale shown beneath it. How these are measured.

Evidence55: 0100
One briefing, one blog post
Everything about what the product does comes from a single Thursday briefing and OpenAI's own release. Fortune filed the same text twice and PYMNTS quotes the press release directly, so the interface description, the connector count and the roughly 10-minute slide figure are all what OpenAI chose to show. American Banker supplies the only material that did not originate with the vendor: two named practitioners, and a question about eligibility that went unanswered.
Adoption18: 0100
Two design partners, no named buyer
Morgan Stanley and Evercore shaped the feature set; neither is described as paying for the result. Turley said demand was heavy and then declined to name a bank that has signed on. Access runs through an enterprise account plus a clearance conversation, so even the addressable set is whatever OpenAI decides it is -- and that, on launch day, is the whole record.
Hype gap+34: −100 understated0 aligned+100 overstated
Canonical before anyone has bought it
OpenAI calls this "the canonical product we are hoping the industry adopts" and Turley puts it next to Excel's arrival on Wall Street. The supporting record is a vendor demo, a 10-minute slide render, and no identified customer. Anthropic shipped a finance version of Claude 14 months earlier, so the claim to be the industry's one product is a marketing position; nothing observed supports it.
Incentives80: 0100
Announced into the enterprise revenue story
The launch lands while OpenAI is telling investors that enterprise now out-earns consumer, and CNBC reports the company is gearing up for an expected IPO. The product detail reaches readers through a briefing OpenAI convened, a demo it ran and a blog post it wrote. Turley also had a direct reason to answer the junior-hiring question the way he did: a labour-replacement story is harder to sell to the firms that must buy the seats.
Confidence64: 0100
Solid on the what, thin on the whether
What was announced, who helped build it, how access is gated and what Turley said are all firmly sourced across four publishers, two of which quote the same briefing. Output quality, the real cost at high effort, and whether any bank has bought it sit outside what anyone has reported.

Topics and entities

Topics

Enterprise AI Sales
Automation of entry-level work
Vertical AI Products
AI in investment banking