Anthropic Walked Into Excel and Nobody Asked Them To. That's the Point.

The Macro: Spreadsheets Never Needed Saving Until AI Made That Possible

Every analyst I know has a spreadsheet they’re embarrassed about. Twelve tabs, circular references, a formula so long it wraps four lines, and the person who built it left eighteen months ago. This is not a niche problem. This is the default state of financial modeling at most companies.

Excel has roughly 750 million users worldwide. That number gets cited so often it’s started to feel like wallpaper, but stop and actually sit with it. Three quarters of a billion people are doing work inside a grid of cells. A meaningful slice of them are doing genuinely complex work, scenario modeling, cap table management, revenue forecasting, the kind of work where a broken formula costs real money.

The data analytics market is growing fast regardless of which research firm you ask. Fortune Business Insights puts it at $82 billion in 2025 heading toward nearly $500 billion by 2034. Grand View Research says $302 billion by 2030. The specific numbers disagree; the direction does not. There is a large and expanding market of people who need to do more with data than they currently can.

The interesting thing about where AI has landed in this space is that most tools approached it from outside the spreadsheet. You’d paste data into a chat interface, get an answer, paste it back. Tools like Formaly have been working on this problem from different angles. The incumbent move from Microsoft was Copilot, which has had a complicated rollout and mixed reviews from power users.

What Anthropic did with Claude in Excel is different in kind, not just degree. They didn’t build a spreadsheet tool. They put Claude inside Excel itself, as a native add-in, with full workbook awareness. That’s a distribution strategy as much as a product decision, and it’s the kind of move that tends to make category-specific startups nervous.

The Micro: What Full Workbook Awareness Actually Means in Practice

The phrase that matters in the product description is “full workbook awareness.” Not the active sheet. The whole workbook. Tabs, cross-tab dependencies, named ranges, the works.

Most AI integrations for spreadsheets operate at the cell or selection level. You highlight a range, ask a question, get an answer about that range. Claude in Excel, according to multiple LinkedIn posts from people who tested it, reads the entire model before responding. One user reportedly built an eleven-tab financial model in ten minutes and described the experience of having Claude explain the dependencies as genuinely useful rather than superficially impressive.

The design decisions are worth paying attention to. According to user reports, Claude avoids overwriting existing cells by default. It tracks and explains every action it takes. It provides cell-level citations so you can see exactly what it touched and why. These are not flashy features. They are the features that matter when you’re working on something real and you cannot afford to have an AI silently break your model.

It launched in beta, available to Pro Max and enterprise plan subscribers according to LinkedIn posts from people testing it. It got solid traction on Product Hunt when it launched. The comments were sparse but the interest was clear.

One comparison I keep seeing in the wild: people saying Claude in Excel works better than Excel’s own Copilot AI for actual formula work. That’s a remarkable thing to be able to say about a third-party add-in living inside a Microsoft product, and I’d want more systematic testing before I leaned on it hard, but the pattern across multiple independent posts is consistent enough to flag.

For analysts tracking what’s happening across tools and traffic, the same pattern shows up in adjacent categories. You Think You Know Your Web Traffic. You Don’t. The tool that reads more context wins. That thesis seems to be what Anthropic is betting on here.

The Verdict

This is a serious product with a serious distribution advantage.

Anthropathic didn’t build a standalone app and hope people would switch workflows. They put Claude where the work already happens. That’s the correct instinct, and it’s harder to execute than it sounds. The native sidebar approach, the full workbook context, the conservative defaults around cell overwrites, these suggest a team that thought carefully about what would actually make analysts trust the thing.

The real question at 30 days is whether the beta performance holds across messy real-world models, not clean demo spreadsheets built to show the AI at its best. At 60 days, I’d want to know how enterprise IT is responding to a Claude add-in sitting inside Excel with access to full workbooks. That’s a security conversation that will happen, and how Anthropic handles it matters.

At 90 days, the competitive pressure gets interesting. Microsoft is not going to ignore this. They have Copilot and they have the home field advantage. Cube Wants to Be the Semantic Layer AI Actually Uses is a good example of the kind of infrastructure bets being made around exactly this problem, and the space is not standing still.

I think Claude in Excel is genuinely useful right now for the analysts who can access it. The larger story is whether Anthropic turns a useful add-in into a durable position inside enterprise workflows. That’s a harder problem than the spreadsheet one.