OpenAI Puts ChatGPT Inside Excel With GPT-5.4 Built for Finance

OpenAI has launched ChatGPT for Excel in beta, an add-in powered by its new GPT-5.4 model that builds, updates, and audits spreadsheet models directly inside workbooks — and the company claims a near-doubling of performance on its internal investment banking benchmark to back it up.

On that benchmark, which evaluates real-world workflows such as building a three-statement model with proper formatting and citations, performance improved from 43.7% with GPT-5 to 87.3% with GPT-5.4 Thinking, according to OpenAI. That number is the clearest signal of what this launch is really about: OpenAI wants finance professionals — analysts, strategists, researchers, and accountants — to treat ChatGPT as core infrastructure for modeling work, not a chatbot bolted onto the side of it.

GPT-5.4 is available today as GPT-5.4 Thinking in ChatGPT, Codex, and the API. OpenAI describes it as its most advanced model, optimized for financial reasoning and Excel-based modeling. The company says it worked closely with industry practitioners to improve the model on tasks that often take analysts hours or days: financial modeling, scenario analysis, data extraction, and long-form research.

What ChatGPT for Excel actually does

The add-in embeds ChatGPT directly in spreadsheets, where it can build, analyze, and update models using the same formulas and structures teams already rely on. OpenAI's pitch is straightforward: users describe what they need in plain language, and ChatGPT creates or updates live Excel models in the workbook, preserving structure, formulas, and assumptions in a formatted, Excel-native file.

Three capabilities define the beta.

First, model building and scenario analysis. Instead of manual construction, teams can describe a budget, inventory report, or data analysis task and let ChatGPT generate the working model.

Second, reasoning across workbooks. ChatGPT can understand how sheets and formulas connect across a model, explain why outputs changed, trace and fix errors, and show how assumptions flow through. OpenAI flags this as especially useful when analysts inherit existing templates and need to get up to speed quickly before making decisions.

Third, auditability. ChatGPT explains what it is doing as it works and links answers to the exact cells it references and updates. Because calculations run directly in Excel, teams can trace assumptions, audit formulas, and verify how results were produced. Before making changes to a workbook, ChatGPT asks for permission, and users can review each step and undo edits.

That permission-first design matters for the target market. Banks and asset managers operate under audit requirements where unexplained changes to a valuation model are a compliance problem, not just an inconvenience.

Known limitations

OpenAI is candid about the beta's rough edges. Some responses may take longer while the company optimizes performance. Generated outputs may occasionally require cleanup to match preferred formatting or layout conventions. Complex formulas and edge cases may still need manual refinement, even though ChatGPT can generate and explain formulas.

Availability

ChatGPT for Excel in beta is available globally to ChatGPT Business, Enterprise, Edu, Teachers, and K-12 users, and to ChatGPT Pro and Plus users outside the EU. ChatGPT for Google Sheets is coming soon.

In Enterprise, Edu, and Teacher workspaces, access is off by default. Admins can enable it for specific users with custom roles and group permissions — a governance choice consistent with how cautious IT departments prefer to roll out AI tools.

The timeline has already moved since the original announcement. An April 22, 2026 update brought ChatGPT for Google Sheets into beta and added app integrations and skills for both spreadsheet add-ins. A May 5, 2026 update made both generally available across all plans, powered by GPT-5.5.

Financial data integrations

The spreadsheet add-in is half the launch. OpenAI is also shipping financial data integrations directly in ChatGPT for FactSet, Dow Jones Factiva, LSEG, Daloopa, S&P Global, Moody's, MSCI, Third Bridge, and MT Newswires, with FactSet listed as coming soon in the initial release. The integrations bring market, company, and internal data into a single workflow, and OpenAI says the apps help users produce cited outputs — earnings summaries, valuation snapshots, credit memos — faster by cutting time spent gathering inputs.

The company also supports proprietary data through custom apps built on the Model Context Protocol (MCP). With GPT-5.4's longer context and ability to handle more complex tasks, OpenAI says teams can move faster on company research, model refreshes, and cited outputs for valuation, diligence, underwriting, and related work.

For due diligence specifically, teams can use apps alongside ChatGPT's research features to pull from filings, transcripts, decks, and spreadsheets, then produce structured, cited outputs that export to PDF or Microsoft Word. Recent updates added more control over the research process: users can focus on specific websites and data sources, shape the research plan before and during a run, and review sources and citations in a redesigned workspace.

Security and governance

OpenAI is pairing the finance push with an enterprise controls story, which will decide adoption in regulated environments. ChatGPT Enterprise includes RBAC, SAML SSO, SCIM, and audit logs, with support for common DLP and SIEM tools. Data is protected with encryption in transit using TLS 1.2+ and at rest using AES-256, plus enterprise key management support. Data residency and regional processing controls address regional data requirements.

One default matters most to legal teams: data shared with ChatGPT Enterprise is not used to train or improve OpenAI's models.

Why this matters

The launch targets one of the highest-value, most formula-driven professional workflows still done largely by hand. Financial modeling is repetitive, error-prone, and time-consuming — and it sits inside institutions with budgets to match. If GPT-5.4 can reliably build a three-statement model at anything close to the 87.3% benchmark figure suggests, the economics of junior analyst work change.

The competitive stakes are equally clear. Microsoft is embedding AI across its own Office ecosystem, and Bloomberg, FactSet, and S&P Global are building AI into their terminals and platforms. OpenAI's counter is to go where the work happens — the Excel workbook — while wiring in the same data providers those incumbents control. The fact that FactSet, LSEG, and S&P Global appear as integration partners signals that data vendors are choosing to distribute through ChatGPT rather than fence it off.

OpenAI says it is already working with financial institutions deploying ChatGPT across research, underwriting, auditing, client engagement, code modernization, and operations, and is seeing impact in due diligence, client experience, and investment research workflows across banks, asset managers, and insurers.

The company's stated next step is continued iteration based on real-world deployments. Enterprise customers can build directly with OpenAI or work with partners including Accenture, Bain, Boston Consulting Group, McKinsey & Company, and PwC to integrate the tools into existing data, applications, and operating models. With general availability now shipped on GPT-5.5, the open question is whether the benchmark gains hold up in production — and whether institutions trust the audit trail enough to let models, not just analysis, be machine-built.