NEW YORK — The workflow OpenAI demonstrated for its new investment banking plugin looked like a junior banker’s morning compressed into seconds. Codex pulled the comparable transactions, checked them against the filing, drafted the valuation summary, and formatted the output into a memo that could be sent to a managing director. OpenAI did not need to explain why it built this particular plugin. The banking industry already understood the message: Codex is no longer just a tool for programmers.
The company unveiled six profession-specific plugins for Codex this week, aimed at data analysis, creative production, sales, product design, equity investment, and investment banking, alongside two new platform features called Sites and Annotations. The releases complete a repositioning that has been building for a year: Codex, which began as a coding assistant, is being sold as an enterprise productivity platform, with OpenAI positioning it directly against Microsoft’s Copilot for control of the workplace AI market.
The numbers explain the ambition. Codex has roughly five million weekly active users, and OpenAI says about 20 percent of them are knowledge workers rather than software engineers. That is the audience the new plugins target: analysts, designers, salespeople, and deal teams who have never written a line of code but who spend their days producing documents, models, and presentations that an AI agent can help assemble.
The financial services plugins are the most significant, according to analysts. Investment banking and equity investment are among the most process-heavy, document-dense professions in the economy, and they have been slow to adopt AI beyond pilot projects. A plugin that can draft deal memos, summarize filings, screen comparable transactions, and format materials for internal review is not a novelty; it is an entry ticket into workflows that have resisted automation for decades. “This is the first time a general AI platform has gone after the core workflow of finance rather than the back office,” said one analyst who follows enterprise software.
The financial industry is watching closely, and warily. Compliance rules require that any AI output used in client-facing materials be reviewed by humans, and firms are still working out how to document AI involvement in deal processes. OpenAI has addressed the concern with Annotations, a feature that records which parts of a document were generated or edited by the agent, creating an audit trail that compliance teams can inspect. Whether that trail satisfies regulators in banking is an open question, but the feature signals that OpenAI is treating the financial market as a serious, long-term target.
Sites, the other new feature, turns Codex into a publishing tool, allowing users to build and deploy web pages from a prompt. The feature extends Codex’s reach from documents to the public internet, and analysts said it puts OpenAI in competition with the website builders that small businesses and marketing teams already use. For a platform that began as a developer tool, the move is a clear step toward general business software.
The competitive front that matters most runs through Microsoft. OpenAI’s largest backer has spent two years embedding Copilot into the same office workflows the new plugins target, and Microsoft’s Scout assistant, unveiled this week, is designed to manage those workflows on its own. The two companies are now selling competing visions of the same future: OpenAI’s is a platform built around a single strong model family; Microsoft’s is a platform built around the office software where work already happens. Analysts said the overlap has moved from theoretical to direct.
The plugins also carry a pricing message. OpenAI is betting that enterprises will pay premium per-seat prices for profession-specific agents rather than generic assistants, and the banking and investment plugins are the clearest test of that thesis. If a bulge-bracket firm signs a contract for a thousand banking seats, it validates an entire business model. If financial firms treat the plugins as toys for junior staff, OpenAI will have to rethink the pitch.
The sales and product-design plugins extend the same logic to other departments. A sales team can hand Codex a call transcript and receive a follow-up email, a deal summary, and a pipeline update in the customer relationship system. A product designer can describe a feature and get interface drafts, user-flow diagrams, and a spec document in return. These are the repetitive, template-heavy tasks that occupy the middle of the working day, and OpenAI’s bet is that automating them will feel less like replacing a job and more like removing the worst part of it.
For the broader AI market, the significance is directional. Until now, AI adoption in finance has been concentrated in risk, fraud detection, and internal coding. Codex’s banking plugin pushes the technology into deal execution, the revenue-generating core of the industry. Whether the push succeeds depends on the same forces that have governed finance’s slow adoption of every previous technology: cost, compliance, and trust. OpenAI has addressed the first two. The third is earned in production, one deal at a time.
