Microsoft, Google and Anthropic Channel Enterprise AI Use With Spreadsheets

With many financial planning teams still running on Excel despite years of alternatives, AI developers are abandoning the strategy of pulling users to new platforms. Instead, they are bringing the technology to where the work already happens. The thesis is straightforward: spreadsheet-embedded AI eliminates the three most common barriers to adoption: skills gaps, budget constraints and integration complexity.

The Interface Has Already Shifted

Microsoft’s Agent Mode in Excel, introduced in late 2025, moves beyond single-turn Copilot replies into agentic workflows. A model-powered agent plans, executes and validates multi-step spreadsheet tasks within the workbook, building tables, writing formulas and producing charts as it goes. Users input a natural-language prompt, and the agent handles the rest, surfacing each step for review.

Google is pursuing a parallel track. Gemini in Google Sheets now lets users build or edit entire spreadsheets with natural language, synthesizing data across files, emails, chat and the web. Google reported on March 10 that Gemini in Sheets achieved a 70.48% success rate on the full Spreadsheet Bench dataset, which it characterized as nearing human expert ability.

Anthropic is also moving into the same grid. According to a March 11 report from VentureBeat, the company launched shared context across its Claude for Excel and Claude for PowerPoint add-ins, enabling Claude to carry information, instructions and task history between an open spreadsheet and an open presentation in a continuous session.

A financial analyst can pull comparable company financials from a workbook, build a trading comps table in Excel, drop the valuation summary into a pitch deck and draft the summary email without switching tabs or re-explaining the dataset at each step, according to VentureBeat. The update also introduced Skills: reusable one-click workflows that teams save and share organization-wide, covering standardized processes like variance analyses and approved slide templates.

On the AI-native side, Shortcut, built by MIT spinout Fundamental Research Labs and powered by Anthropic’s Claude, functions as a system of AI agents that completes complex Excel tasks, including full discounted cash flow models, from a single natural-language prompt. The tool scored over 80% on Microsoft Excel World Championship benchmark cases and finishes them about 10 times faster than human analysts, according to PYMNTS.

Each of these products uses the same delivery strategy: bringing AI into the spreadsheet rather than pulling users away from it.

The Zero-Friction Onramp

The three leading barriers to enterprise AI adoption are a lack of in-house skills, insufficient budget and integration complexity. Spreadsheet-delivered AI eliminates all three at once. Microsoft 365 Copilot and Google Workspace with Gemini bundle AI into existing subscription tiers. Deployment becomes a settings toggle, not a procurement event.

That distribution advantage compounds because users never left the spreadsheet in the first place. A January 2025 survey by the Association of Financial Professionals found that 96% of FP&A teams still use Excel for planning despite years of purpose-built alternatives. Any AI tool requiring migration faces the installed-base inertia. Any AI tool delivered within the spreadsheet is included for free.

The stickiness runs deeper than habit. An analysis from a16z’s speedrun newsletter draws a useful distinction between two use categories: “mini software” workflows like dashboards and trackers, which AI may eventually replace, and analytical workflows like financial models, where building the spreadsheet itself is how practitioners reason through the inputs. The second category dominates finance, strategy and operations. For those users, the spreadsheet is not storage. It is the reasoning surface. AI augmentation inside the grid fits. Replacement tools do not.

Anthropic has moved inside the same layer. The trajectory from a single Excel plugin to a shared cross-app context to an embedded engine inside Microsoft Copilot gives Claude enterprise distribution that a standalone app cannot replicate.

The hardest part of enterprise AI deployment has never been the model. It has been getting employees to use it inside real workflows without a dedicated implementation effort. Integrating AI in spreadsheets solves that.

For all PYMNTS AI and B2B coverage, subscribe to the daily AI and B2B Newsletters.