How to Set Up ChatGPT for Excel (2026 Add-In)

ChatGPT for Excel is a native Excel add-in, not an external tool you toggle between. Install it once from the Microsoft Add-ins store and a ChatGPT panel appears inside your workbook. From there, you describe what you want in plain English and the add-in builds or modifies formulas, summarizes data, and updates models directly in your cells.

It launched on March 5, 2026, powered by GPT-5.4. On OpenAI's internal investment banking benchmark, GPT-5.4 scored 87.3% accuracy compared to 43.7% with the prior GPT-5, a meaningful jump for complex financial modeling tasks.

Who Can Use It

The add-in is available to ChatGPT Pro, Plus, Business, Enterprise, Edu, Teachers, and K-12 account holders. Pro and Plus users outside the European Union are included. EU users are not currently supported due to data regulation compliance work still in progress.

Free ChatGPT accounts and Team plan users do not have access yet. You can verify eligibility at chatgpt.com/apps/spreadsheets.

Enterprise and Edu workspace admins must enable the add-in for users before it appears. It is off by default in those environments. Admins can set granular permissions by user group through the admin console.

How to Install the Add-In

The install process takes under two minutes.

Open Microsoft Excel on desktop or in the browser. Go to the Home tab and click Add-ins. In the search bar, type ChatGPT. Select the OpenAI ChatGPT add-in from the results and click Add.

Once installed, a ChatGPT button appears in your ribbon. Click it to open the sidebar panel. Sign in with the same OpenAI account that holds your eligible plan.

The add-in works in both Excel desktop (Windows and Mac) and Excel on the web. The feature set is largely the same across both environments. One exception: VBA macro generation is only available on desktop Excel, since Excel on the web uses Office Scripts instead.

What You Can Do With It

Build and update formulas. Describe the output you need and ChatGPT generates the formula and inserts it into the target cell. It handles lookup functions (XLOOKUP, INDEX/MATCH), conditional aggregation (SUMIFS, COUNTIFS), data cleaning (TRIM, SUBSTITUTE, PROPER), and date calculations.

A practical example: "In column D, calculate the number of business days between the date in column B and today." ChatGPT writes the correct NETWORKDAYS formula and applies it down the column.

Reason across multiple sheets. The add-in reads the full structure of a multi-sheet workbook and traces how data flows between sheets. If a summary sheet is pulling incorrect totals because a reference broke, describe the problem and ChatGPT identifies the broken link and suggests a fix.

Analyze and summarize data ranges. Highlight a table, open the panel, and ask questions directly. "What are the top five customers by revenue this quarter?" or "Which product category has the highest return rate?" ChatGPT returns the answer and, where useful, generates a formula that surfaces that result dynamically.

Build scenario models. Finance and ops teams can describe a scenario and ask ChatGPT to build the supporting logic. "Add a section that shows what happens to gross margin if COGS rises by 10%, 15%, or 20%." The add-in inserts the rows, writes formulas that reference existing assumptions, and formats them consistently with the surrounding model.

Explain existing formulas. Paste a complex formula into the panel and ask for a plain-English breakdown. This is useful when inheriting a workbook or auditing someone else's model before a presentation.

What It Cannot Do Yet

The add-in is in beta and several capabilities are out of scope for now.

Power Query, Pivot Tables, and Data Model connections are not accessible to the add-in. If your reporting relies on Power Query refreshes or pivot slicers, ChatGPT cannot modify those components directly.

VBA macros are partially supported. The add-in can suggest VBA code in the chat panel that you paste into the VBA editor yourself, but it cannot create or run macros autonomously.

Office Scripts (the web equivalent of VBA) are similarly unsupported for direct execution from the add-in.

Slicers, timelines, the named ranges manager, and external data connections are outside the current scope. Very large workbooks with hundreds of thousands of rows may also hit context window limits, causing partial results or prompts to work with a smaller sample.

Data Privacy Before You Start

Account type determines how your data is handled inside the panel.

On Business and Enterprise plans, OpenAI does not train on your inputs by default and workspace admins control data handling policies. On Pro and Plus plans, the default settings allow OpenAI to use conversation data for model improvement unless you opt out under Settings > Data Controls.

For workbooks containing customer personally identifiable information, financial projections, or proprietary pricing, mask or anonymize the data before pasting it into the panel. Sample rows work well for testing the add-in's capabilities before running it on production data.

A practical precaution: duplicate the workbook before letting ChatGPT make changes, especially during beta. This gives you a clean rollback point if the add-in modifies cell references or formatting in unexpected ways.

A Practical Workflow for Analysts

Users who tested the add-in in the weeks after launch report the most reliable results when they work in focused, single-step instructions rather than sending long multi-part requests.

Start specific. "Format column C as currency with two decimal places" works better than "format all financial columns, add a total row, and highlight anything over $10,000." After each change, verify the output before moving to the next instruction.

For formula auditing, highlight the range in question, open the panel, and ask ChatGPT to explain the logic. Then follow up with "identify any cells that could produce errors if the input data changes format."

For report generation, describe the output structure first. "I want a summary table showing revenue by region, with month-over-month change and a YTD total column." Once ChatGPT builds the skeleton, add formatting and data requests in separate follow-up messages.

If your raw data lives outside Excel (a database export, a CSV from your CRM, or an API pull), cleaning and structuring it before it reaches Excel saves time. Tools like VSLZ are built for this upstream step and let you clean and analyze a file from a single prompt before the data moves into your model.

Summary

ChatGPT for Excel delivers real value for analysts and ops teams who live in spreadsheets. Installation takes two minutes, formula and modeling assistance handles a meaningful share of day-to-day work, and cross-sheet reasoning works better than most users expect.

The current beta limitations are meaningful: no Power Query, no Pivot Table modification, no autonomous macro execution. Work within those constraints and the add-in cuts through a significant portion of the repetitive formula and analysis work that currently takes hours.

Eligible accounts can access the add-in now at chatgpt.com/apps/spreadsheets.