NextFin News - OpenAI has taken direct aim at the most labor-intensive layer of Wall Street, unveiling a version of ChatGPT built to do the research, data analysis and pitchbook formatting that junior investment bankers have spent decades learning to do by hand. The product, called ChatGPT for Financial Services, is a tailored edition of ChatGPT Work powered by GPT-6 Astra and developed with design partners Morgan Stanley and Evercore, Nick Turley, OpenAI's vice president of product, said at a briefing on Thursday.
The launch lands at a pivotal moment for OpenAI. Chief financial officer Sarah Friar told investors in mid-August that the company's enterprise business had overtaken its consumer business for the first time, with annualized revenue running at roughly $40 billion — double the pace at the start of the year. A sector-specific product for finance is the clearest signal yet that OpenAI is now selling itself less as a consumer chatbot and more as infrastructure for the world's most data-heavy industries.
The central tension is straightforward: OpenAI is pitching the tool as a productivity multiplier, but the work it automates is the same work that Wall Street has used for half a century to train its next generation of dealmakers. If the apprenticeship layer disappears, the industry's pipeline disappears with it — and that is a problem no amount of efficiency solves.
What the product actually does
The pitch is concrete rather than aspirational. In a live demonstration, Turley showed the system analyzing a potential merger target: pulling financial figures from industry-standard data sources, selecting comparable companies, building a chart and exporting a PowerPoint deck formatted to a bank's pre-existing style guide.
"It's very easy to make slides that look good, but it's much harder to make slides that actually make sense," Turley said. "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 selloff and the rebound."
What separates this from the generic ChatGPT Work product is native access to financial data. The system connects directly to LSEG, Daloopa and PitchBook for financial statements and earnings transcripts, and can tap into a user's firm's existing data subscriptions automatically. Additional datasets from Crunchbase and Quartr cover company fundamentals, with the data indexed on OpenAI's own infrastructure to improve retrieval and citation. The product also produces citations that let an analyst trace a number back to its source filing, lets users audit the charts it builds, and adds administrative controls for sensitive deal materials — a direct response to the confidentiality concerns that have kept banks from putting live transactions into general-purpose AI tools.
OpenAI's stated ambition is to teach the model to work like a trained analyst rather than merely summarize text.
"We're effectively teaching ChatGPT to research like an analyst and back up its conclusions like an analyst as well," Turley said.
The initial focus is investment banking and equity research, though Turley said the company plans to release tailored versions for "a number of sectors" beyond financial services. He declined to name which banks have signed on, but the choice of design partners is itself a signal: Morgan Stanley and Evercore span both the wealth-management and the pure advisory ends of the business, giving OpenAI a test bed across two very different workflows.
The apprenticeship model under pressure
The product arrives with a question Wall Street has been avoiding: if the tasks that consume an analyst's first years — company research, comps, pitchbook formatting — can be executed in minutes, what happens to the industry's apprenticeship model? Investment banks have long justified 100-hour weeks for recent college graduates as the price of training the next generation of dealmakers. The implicit bargain is simple: juniors trade their time for tacit knowledge, and the industry trades training costs for a reliable supply of future vice presidents and rainmakers.
Turley framed the tool as an efficiency gain rather than a headcount cut, comparing it to the spreadsheet revolution.
"If you study the life of an analyst or of a banker, depending on the industry, they're working 100-hour weeks," Turley said. "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 Excel analogy is flattering — and incomplete. Spreadsheets automated calculation but left the reasoning to the banker. ChatGPT for Financial Services automates a portion of the reasoning itself: choosing peers, reconciling a chart to underlying data, explaining a selloff. That is precisely the work through which junior bankers historically learned how deals are valued, how arguments are structured and where numbers come from. When the tool checks the chart against the data and explains the selloff, it is not just saving time; it is performing the act of verification that used to be the analyst's education.
One month before OpenAI's launch, Chris Churchman, the Goldman Sachs partner who runs one of the bank's flagship AI projects, warned that automating exactly these tasks risks "cognitive atrophy" in the next generation of financiers.
"There's a huge danger here that in the era of AI, we outsource our reasoning to these models, and we have cognitive atrophy that stops us being able to reason from first principles ourselves," Churchman said in an episode of Goldman's Exchanges podcast.
Churchman's warning was specific about what gets lost. "You still need to reason about problems and structure it into an argument, and now we're delegating reasoning," he said. He also said Goldman has not yet determined how it will manage that transition, and that employees — not automated systems — should make the final call on consequential decisions. That last point matters: a tool that produces a finished argument is far more seductive than one that merely retrieves data, and the default human behavior is to ratify rather than re-derive.
Why finance, and why now
The timing is not accidental. Finance is one of the few industries where the core work product is already digital, structured and citation-heavy, and where the cost of a hallucinated number is immediate and expensive. A sector-specific product lets OpenAI bundle the data access, audit trails and compliance controls that banks require, while charging a premium that a generic chatbot cannot command.
It also reflects OpenAI's rapidly shifting revenue mix. At the start of 2026 the company said enterprise made up roughly 40 percent of revenue and was on track to reach parity with the consumer business only by year-end. That crossover arrived in August, about two quarters early, against a revenue base that had doubled to a $40 billion annualized pace. Business revenue alone rose 32 percent in a single month over the summer. With a confidential IPO filing opened in June and a March funding round that valued the company at $852 billion, OpenAI has a strong incentive to prove that its enterprise business can keep compounding — and that it can lock in high-value verticals before rivals do.
The broader internal signal is even more telling. Days before the financial-services launch, OpenAI disclosed that it had hit an internal milestone it set a year earlier: an "automated research intern" capable of completing well-defined research tasks under human direction that would take a skilled researcher several days. Inside OpenAI's own research organization, agent runtime had flipped past human labor — 3.1 agent-workdays for every human workday as of mid-August. The company's next internal target is a fully automated AI researcher by March 2028.
That timeline matters for Wall Street because the same capability curve applies to any knowledge-work firm that bills by the hour. The first task AI absorbs is not the final judgment — humans still choose research priorities and decide whether systems should be scaled, paused or deployed. It is the apprenticeship layer beneath the judgment: the retrieval, the reconciliation, the first draft. Finance is simply the first vertical where that layer is expensive enough, structured enough and citation-sensitive enough to productize.
Competition is already in the building. Anthropic announced its own tailored solution, Claude for Financial Services, last year, and the race for Wall Street's AI budget is now a fight over workflow integration rather than raw model capability. The winner will be the product that sits inside the bank's existing data subscriptions and document pipelines, not the one with the highest benchmark score. OpenAI's data partnerships — LSEG, Daloopa, PitchBook, Crunchbase, Quartr — are a moat attempt: once a bank's analysts are querying live deal data inside ChatGPT, switching costs rise quickly.
The second-order effect nobody is pricing in
The obvious first-order read is that junior bankers will spend less time on grunt work and more time on judgment. The second-order effect runs deeper and cuts the other way: the economic justification for large analyst classes weakens at the same time that the training pipeline is hollowed out. Banks do not hire analysts because they love training; they hire them because the apprenticeship model has historically produced a reliable supply of associates, vice presidents and eventual rainmakers. If AI absorbs the repetitive layer, banks face a choice between shrinking entry-level cohorts — and accepting a thinner bench in five years — or redefining what a first-year banker is for.
There is a precedent for how this plays out, and it is not comforting for the middle of the profession. Investment banking compensation and prestige have always concentrated at the top and the client-facing edges; the middle layers have been the first to go in every restructuring. AI does not change that distribution — it accelerates it downward. The analysts who remain will be pushed toward judgment and client work earlier, which sounds like an upgrade until you ask who trains them to do it.
That tension is already visible in the language each side uses. OpenAI speaks of productivity per employee; Goldman speaks of preserving institutional knowledge. Both can be true at once: headcount at the bottom can fall while the value of the remaining humans rises, concentrated in judgment, client relationships and the final sign-off. The question is whether the smaller cohort that survives produces enough future leaders to staff the partnership track a decade from now.
The counter-thesis
The strongest argument against the displacement view is historical. Investment banks have absorbed every labor-saving technology since the spreadsheet without shrinking their analyst ranks, because cheaper analysis simply expanded the volume of deals, pitches and coverage. If AI makes it materially cheaper to produce a pitchbook, banks may respond by pitching more clients, covering more companies and staffing more live transactions — absorbing the productivity gain into volume rather than cutting headcount. Under that reading, the number of junior bankers stays roughly flat while their output per hour rises sharply, and the real beneficiaries are the banks' margins.
That scenario is plausible, but it rests on one condition: deal flow must be elastic enough to absorb a step-change in productivity. If M&A volumes do not expand proportionally, the labor math changes. The falsifying signal is straightforward — watch analyst and associate hiring cohorts at the bulge-bracket banks over the next two recruiting cycles. If headcount holds steady or grows while AI tooling rolls out broadly, the volume-absorption thesis wins. If cohorts shrink materially while pitch volumes rise, the displacement thesis is confirmed.
Scenarios: what happens next
The base case is complementarity with gradual compression. Banks adopt the tool for research and formatting, analyst productivity rises, and headcount grows more slowly than revenue rather than falling outright. The apprenticeship model adapts unevenly — some firms redesign training around AI-supervised work, others let the tool do the teaching and accept the atrophy risk Churchman flagged.
The upside case for the technology is a renaissance in coverage. If AI cuts the cost of a pitchbook by half or more, mid-sized companies that were previously too small to pitch become viable targets, and equity research coverage expands into neglected names. Productivity gains translate into volume, headcount holds, and the industry grows its way out of the problem. This is the world Turley is describing when he invokes Excel.
The downside case is a hollowed pipeline. If deal volumes stay muted and banks respond to margin pressure by shrinking entry classes, the industry gets a short-term profit pop and a long-term staffing gap. Five to seven years from now, the associate and vice-president benches are thinner, and the firms that invested in human training while their peers automated it hold a scarce-asset advantage. The irony is that the banks that resist the tool most aggressively may be the ones best positioned a decade out — not because the tool is bad, but because the training it replaces is more valuable than the time it saves.
What to watch
In the near term, the signal to track is adoption: which banks sign on, how many seats they deploy, and whether the tool moves from pilot to daily workflow inside live deal teams. Medium term, the key data point is the composition of entry-level hiring classes at the largest banks — the clearest read on whether AI is complementing or replacing the apprenticeship pipeline. Longer term, the question is whether finance becomes a template for OpenAI's vertical strategy: Turley has already signaled tailored products for other sectors, and the same playbook — proprietary data access, citation trails, compliance controls — can be replayed in law, consulting and healthcare.
The central judgment: OpenAI is not just selling a tool to bankers; it is selling a wedge into the highest-value enterprise workflows in the economy, and finance is the proof of concept. Whether that wedge hollows out the junior ranks or simply makes them more productive will be decided not by the technology but by whether deal volume can grow fast enough to absorb it. The spreadsheet made bankers faster without making them fewer. This time, the tool thinks as well as it calculates — and that difference is what the apprenticeship model cannot survive unchanged.
Explore more exclusive insights at nextfin.ai.
