Claude in Excel: The $80/Month Feature That's Now $20

The Excel AI sidebar that used to be a premium-tier perk now rides along with the plan most marketers already pay for.

For a while, Anthropic’s Excel add-in was locked behind the top Max tier, the one that runs well into three figures a month. Then it opened up to the regular Pro plan, the roughly $20-a-month tier most marketers already carry. Same tool, a fraction of the cost. If you live in spreadsheets, building campaign models, picking apart performance data, scrubbing CRM exports, that change matters.

I first wrote this in January, when the add-in was still a preview and the rules were fuzzy. It’s a different product now. Here’s what changed, what it does, where it earns its keep, and the spots where it still falls short.

Updated August 19, 2026. Claude for Excel went generally available on May 7, 2026, alongside Word and PowerPoint. Three things I told you in January are no longer true: it does not run on Excel 2016, it does not lock you to a single Opus model, and it can now build conditional formatting and data validation. All corrected below.

Claude in Excel: Why the $20 Plan Changes EVERYTHING (MarketingAlec on YouTube)

What Claude in Excel actually is

Claude in Excel is Anthropic’s official Microsoft Excel add-in. It drops a sidebar into your Excel window where you talk to Claude about the workbook you have open.

This isn’t copy-paste or a browser extension bolted on the side. Claude reads your open workbook directly: cells, formulas, tab structure, all of it. When it explains something, it cites the exact cells it’s looking at, and you can click a citation to jump straight to that cell. When it changes something, it warns you before it overwrites anything. You can see the work.

As of August 2026, here’s the accurate spec sheet:

• Status: generally available since May 7, 2026 — no longer a preview. Word and PowerPoint shipped GA the same day; Outlook is still in beta.
• Plans: Pro, Max, Team, and Enterprise. The Free plan does not get it.
• Platforms: Excel on the web; Excel on Windows with a Microsoft 365 subscription (build 16.0.13127.20296 or later); Excel on Mac (version 16.46 or later, build 21011600 or later).
• Model: a curated picker, not one fixed model. The add-in shows a shorter list than claude.ai, and your organization’s model-access settings apply on top.
• Install: Microsoft AppSource, “Claude for Microsoft 365.”

The correction on Excel versions

In January I wrote “Excel 2016 and newer.” That was wrong, and it’s the single most likely reason someone tells you the add-in doesn’t work.

Anthropic explicitly lists these as unsupported:

• Excel 2016 and 2019 perpetual or volume license
• Excel on iPad (the add-in needs SharedRuntime, which iPad doesn’t provide)
• Excel on Android
• Older Microsoft 365 builds below the SharedRuntime threshold

A Microsoft 365 subscription is the requirement. A perpetual license with a big version number on it is not the same thing.

The pricing shift

The short version, and the reason this post exists: Excel access landed on Pro, the ~$20/month plan. Before that it was Max-only. That’s what took it from a “maybe later” tool to one I’d recommend to a marketer on a normal plan.

Current list pricing, verified on Anthropic’s pricing page on August 19, 2026:

There’s no separate charge for the add-in and no Microsoft surcharge. It authenticates against the Claude subscription you already have.

What’s new since January

If you tried it in the winter and shrugged, it’s worth another look. The list of things it couldn’t do got noticeably shorter.

Native Excel operations. It can now sort, filter, edit pivot tables, apply conditional formatting, and create data validation dropdowns. In January it would explain how to do those and leave you to it. Now you just ask.

Chat history persists. It keeps your 50 most recent conversations per app, stored locally in your browser via IndexedDB. Worth knowing the shape of that: it’s device-local, it doesn’t sync, and on Excel for the web the browser can evict it on its own schedule. Don’t build a workflow that depends on the history being there tomorrow.

Persistent Instructions. Settings in the sidebar has an Instructions field — standing preferences that apply to every Excel conversation. “Format numbers with thousand separators.” “Always bold column headers.” Instructions you set in Excel apply only to Excel; Word and PowerPoint keep their own.

Skills. Type / in the sidebar to invoke one directly, or let enabled Skills apply automatically as you work.

Connectors. It can pull outside context into the workbook through connectors — S&P Global, LSEG, Daloopa, plus custom remote-MCP connectors your org sets up.

Cross-app context. One conversation can span your open workbook, deck, doc, and inbox. The toggle is “Let Claude work across files,” and the default differs by plan: on for Pro and Max, off for Team and Enterprise, and it’s per-device. If you’re on a Team seat and cross-app isn’t working, that’s why.

Auto-compaction. Long conversations get compacted automatically so you don’t run out of context. Useful, with a real cost — see the failure modes below.

Where it earns its keep for marketers

This is the part that matters for the AI marketing work I do every week. Here’s where Claude in Excel pulls real weight.

Campaign performance analysis. Feed it a performance export and ask plain questions:

• “Which channels had the highest CPA increase month over month?”
• “Flag campaigns where CTR dropped more than 20% while spend went up.”
• “Build a summary table of my top 10 campaigns by ROAS.”

It points back to specific cells in every answer, so you can check its math instead of trusting it.

Budget modeling. When you’re building a budget with multiple scenarios:

• “Create a budget template with columns for Q1 through Q4, a row per channel, and formulas for totals and year-over-year change.”
• “Flex the growth rate from 5% to 10% and show me the impact on the annual total.”

It updates values while keeping formula relationships intact, so downstream cells recompute correctly.

Data cleaning. For the messy CRM and campaign exports that eat your afternoon:

• “Standardize the dates in column C to YYYY-MM-DD.”
• “Split the Full Name column into First Name and Last Name.”
• “Find and flag duplicate email addresses.”

This is the capability every independent review lands on. It’s the least glamorous thing the tool does and the one that saves the most time.

Reporting templates. For the reports you rebuild every month:

• “Add conditional formatting that turns any cell below target red.”
• “Create a data validation dropdown in column B with our five campaign types.”
• “Add INDEX-MATCH formulas that pull campaign names from the reference sheet.”

Debugging someone else’s spreadsheet. The underrated one:

• “Walk me through how the number in C42 is calculated.”
• “Find the source of the #REF! error in the summary tab.”
• “What assumptions drive this forecast?”

Inheriting a model from someone who left the company used to be a lost afternoon. Now it’s a conversation.

MA free resource: Claude in Excel for Marketers — the full guide, with the prompts, the verification checklist, and the three ways to put Claude on a spreadsheet.

Claude vs Copilot: which one

This is a routing call, the same kind I work through across all my AI tool comparisons. The split has sharpened since January.

Price. Claude Pro is $20/month, and a Claude Team seat is $20/seat/month annual. Microsoft 365 Copilot is $30/user/month annual, and that sits on top of a qualifying M365 license. Claude is roughly a third of the cost.

Where your file lives. Claude works on the workbook you have open, local or cloud, no forced AutoSave. Copilot’s Agent Mode wants the workbook saved to OneDrive or SharePoint. If you work with client data you’d rather keep off the corporate cloud, that’s not a small detail.

Python. Copilot can run Python — pandas, matplotlib — in Microsoft’s cloud containers and drop results back in the sheet. Claude can’t. For heavy statistical work, that’s a real Copilot advantage.

Auditability. This is the one that decides enterprise deployments, and it goes the other way. Copilot inherits the Microsoft Purview compliance estate. Claude for Excel does not appear in Enterprise audit logs or the Compliance API, and it doesn’t inherit your org’s custom data-retention settings. More on that next.

ChatGPT still has no first-party native Excel add-in comparable to either.

Honest read: if you’re a marketer or a small team, Claude at $20 is the better reasoning engine on messy multi-tab work and costs a third as much. If you’re an enterprise with a compliance officer, the governance gap is a genuine obstacle, and the two aren’t mutually exclusive.

The governance gap nobody mentions

I skipped this in January. That was a mistake, because it’s the thing that will stop a rollout cold.

It’s not in your audit logs. Anthropic states it plainly: Claude for Excel activity is not included in Enterprise audit logs or the Compliance API, and it does not inherit custom data-retention settings. If you want an audit trail, you stand up your own OpenTelemetry collector — and note that a custom collector receives spans unfiltered, including prompt content and document references. Treat that endpoint like it holds your data, because it does.

Consumer plans and commercial plans are not the same product legally. Free, Pro, and Max run on consumer terms, where conversations can be used for model training unless you turn that off in privacy settings. Team, Enterprise, and the API run on commercial terms and are excluded from training by default. Anthropic’s SOC 2, ISO 27001, ISO 42001, and HIPAA BAA coverage is scoped to the commercial products — Free, Pro, and Max are explicitly outside it.

The practical translation: a $20 personal Pro seat is a great way for you to evaluate the tool. It is not the right way for a company to put real financial data through it. That’s a Team or Enterprise conversation.

Server-side retention: inputs and outputs are deleted on the backend within 30 days of receipt or generation, with a short cache window after so you can still reach context from a recently closed workbook.

Where it still falls short

What it won’t build:

• Data tables
• Macros and VBA operations

Anthropic’s own “not recommended for” list, which is worth reading as written:

• Final client deliverables without human review
• Audit-critical calculations without verification
• Models containing highly sensitive or regulated data without proper controls

A few workflow notes:

• It only sees workbooks that are currently open. It can’t create, open, close, or switch files on its own.
• No scheduled or unattended runs from inside the add-in.
• Auto-compaction keeps long sessions alive by summarizing away detail, and it doesn’t tell you what it dropped. In a long session, the reasoning behind an early decision may simply be gone. Capture decisions somewhere outside the chat.
• Anthropic publishes no hard workbook size, row, or cell ceiling. Performance degrades on very large models well before you hit any wall, so test your own worst-case file rather than trusting a number you read somewhere.

If you’re maintaining macro-heavy reports with custom VBA, Claude isn’t the tool for that job.

One warning worth taking seriously

Anthropic puts it plainly: only use Claude for Excel with trusted spreadsheets.

That’s not legal-page filler. The risk has a name: prompt injection. Instructions can be hidden inside cells, formulas, or comments. Anthropic says its own testing “identified scenarios where Claude for Excel can be manipulated to extract sensitive information, modify critical data, or perform destructive actions if allowed to act without verification.”

Open an untrusted spreadsheet, ask Claude for help, and buried instructions could push it into:

• Sending data out to an external URL
• Rewriting formulas in ways that mask errors
• Taking destructive actions you didn’t ask for

Claude asks you to confirm before risky operations. Read those confirmations, especially on external files.

So keep it simple:

• Don’t point it at downloaded templates from sources you don’t know
• Be careful with vendor-supplied spreadsheets
• Skip files from email attachments you haven’t checked
• Stick to workbooks you or your team built — or paste the outside data into a workbook you control

That last one is the practical version of the rule. If a file came from outside your company, copy the data into your own workbook rather than handing Claude the file.

Quick start

Ready to try it?

1. Confirm your plan. Pro, Max, Team, or Enterprise. Free doesn’t include it.
2. Confirm your Excel. Microsoft 365 subscription on Windows or Mac, or Excel on the web. Not a 2016/2019 perpetual license.
3. Install the add-in. Microsoft AppSource, search “Claude for Microsoft 365.”
4. Open Excel. Home > Add-ins on Windows, or Tools > Add-ins on Mac. Find Claude.
5. Sign in with your Claude credentials.
6. Start on a copy of a file you built yourself.
7. Ask before you edit. Have it explain the workbook first, then let it make changes.
8. Click the citations. The cell-level citation is the whole verification story. If you don’t click through, you’ve given up the only check the tool offers you.

If you’re rolling this out to a team, your admin deploys it from the Microsoft 365 Admin Center: Settings > Org Settings > User owned apps and services > “Let users access the Office Store,” then Settings > Integrated apps > Add-ins. One gotcha to hand your IT lead: if your org uses Microsoft Entra Privileged Identity Management for admin roles, the Integrated apps page doesn’t recognize PIM-activated roles and deployment fails. It’s a known Microsoft issue, tracking ID 11126536. Deploy from an account with the role permanently active instead.

Go deeper on your own spreadsheets

Two guides, both free, both built around the same rule: Claude builds the analysis, you own the number.

• Claude in Excel for Marketers — the prompts, the verification checklist, and the three ways to put Claude on a spreadsheet. Campaign exports, budget models, CRM lists, attribution.
• Claude in Excel for Construction — for anyone whose spreadsheets go to a bank or a surety. Bid leveling, forecasting, job cost, and the verification discipline that has to come with them.

Go deeper

• Claude for Excel: I’m Obsessed With This New Claude Feature
• AI Is Finally Good at Spreadsheets
• Testing Claude for Excel

Get the spreadsheet prompts that actually pull weight

Every Friday I send marketers one real workflow I’m running, broken down to the prompts and the receipts. The Excel sidebar is one of them, and there’s a stack of others I’d hand a new hire. If you’d rather steal a tested setup than build it from scratch, this is the rest of the kit.

Subscribe free →