Anthropic Extends Claude Across Excel and PowerPoint With Shared Context and Reusable Skills

Anthropic is deepening its push into enterprise workflows by extending its Claude AI model more tightly into Microsoft Excel and PowerPoint. With new shared-context capabilities and a system of reusable “Skills” for repeatable workflows, the company is positioning Claude as a serious option for organizations that want AI to operate inside — not outside — their core Office tools.

The update, available starting March 11 for paid Claude users on Mac and Windows, also broadens how enterprises can deploy Claude through existing cloud AI gateways, reflecting a clear focus on flexibility, governance, and fit with current IT environments.

What Anthropic Is Adding to Excel and PowerPoint

The latest release is centered on two upgraded add-ins: Claude for Excel and Claude for PowerPoint. Both are already available to paying Claude customers, but Anthropic is now changing how they interact and how organizations can standardize work inside them.

First, Anthropic is enabling shared context between the two apps. In a new beta experience, Claude can maintain a continuous session spanning an open spreadsheet and an open presentation. That means the model can remember instructions, data references, and task history across both Excel and PowerPoint without users needing to copy-paste or restate prompts.

Second, Anthropic is introducing Skills — defined, reusable workflows that can be triggered from within the Excel and PowerPoint sidebars. These Skills allow teams to codify common processes and make them available as one-click actions, either personally or across the organization.

Together, these capabilities push Claude beyond a traditional chat assistant toward something closer to an embedded work system: aware of live documents in Excel and PowerPoint, and able to execute standardized, repeatable patterns of analysis and content creation.

Shared Context: One Continuous Session Across Excel and PowerPoint

The shared context beta is the most visible practical change for end users. Starting March 11, paid Claude users on Mac and Windows can open an Excel workbook and a PowerPoint deck and treat them as part of a single continuous session with the model.

In that session, Claude can “carry information, instructions and task history” between the two applications. Concretely, the model can read a live spreadsheet, generate formulas or analysis, and then immediately apply the resulting data or narratives into slides — without users having to restate what data they are working with or manually transfer intermediate outputs.

Anthropic’s own example underscores the workflow focus: a financial analyst can ask Claude to pull comparable company financials from an open workbook, build a trading comps table in Excel, drop the valuation summary into a pitch deck, and then draft the email to a managing director. The key is that this all happens without “switching tabs or re-explaining the dataset at each step.”

This capability builds on Anthropic’s earlier move into Excel: a Claude plugin launched in October 2025 targeted finance professionals with spreadsheet-intensive use cases. The new shared context extends that thinking to multi-application flows, allowing Excel analysis to feed PowerPoint storytelling seamlessly.

For enterprise IT and operations leaders, the change is less about a singular feature and more about a shift in interaction model. Instead of isolated AI tasks per document or app, Anthropic is moving toward session-oriented workstreams that mirror how people already move between spreadsheets and slide decks.

Skills: Turning Tacit Know-How into Repeatable Workflows

The second major element of the release is Skills — a way to formalize and reuse workflows inside the add-ins themselves.

With Skills, users and teams can build and save workflows directly in the Claude sidebars in Excel and PowerPoint. Rather than repeatedly uploading the same references or rewriting similar prompts, organizations can encode standard analyses, review steps, or presentation patterns into one-click actions.

Anthropic frames this as a way to capture processes that previously “lived in one person’s head” and make them available to everyone. A Skill might represent a specific variance analysis, a standard style for investment decks, or a routine data-cleaning operation that a particular team runs every week. Once saved, that Skill can be invoked repeatedly, with Claude executing the defined steps consistently.

Anthropic draws an analogy to its MCP connectors: each Skill, whether scoped to an individual or to the organization, behaves inside the add-ins in a similar, connector-like way. This positions Skills as a building block for structured, semi-automated work rather than ad hoc prompting.

The company distinguishes Skills from Instructions, another configuration layer. Instructions set persistent preferences — for example, number formatting rules in Excel or guidelines for tone and structure when drafting PowerPoint slides. Skills, by contrast, bundle together full workflows to be run on demand.

For leaders trying to standardize how AI is used across teams, this separation is notable. Instructions help enforce global norms; Skills capture repeatable procedures that can be shared, audited, and iterated on like any other operational artifact.

Preloaded Starters: Finance-Heavy Use Cases Out of the Box

To reduce friction for first-time users, Anthropic is shipping a starter set of Skills tuned to common office work — especially in finance and strategy contexts.

On the Excel side, preloaded Skills include:

• Auditing models for formula errors
• Populating DCF and LBO templates
• Cleaning messy data ranges

In PowerPoint, initial Skills focus on:

• Building competitive landscape decks
• Reviewing investment banking materials for narrative alignment

These presets reflect where Anthropic has already seen traction — finance teams, dealmaking workflows, and data-heavy presentations. They also signal the company’s intent: not only to assist with generic office tasks, but to shoulder specific, domain-shaped work where consistent process and quality are critical.

From an adoption perspective, preloaded Skills can serve as reference patterns. Enterprise users can inspect how they are structured, then adapt or extend them to internal standards, creating a library of organization-specific workflows over time.

Enterprise Deployment Options and Governance Considerations

Anthropic is also expanding the ways organizations can deploy these Office integrations, aligning with existing cloud and compliance strategies rather than forcing a single path.

Claude for Excel and Claude for PowerPoint can now be accessed either:

• Through a standard Claude account, or
• Via an existing LLM gateway that routes to Claude models running on Amazon Bedrock, Google Cloud Vertex AI, or Microsoft Foundry.

For enterprise IT teams, this dual approach matters. Many organizations are standardizing on cloud AI platforms that centralize routing, observability, and policy enforcement for LLMs. Allowing the add-ins to plug into Claude through those gateways lets companies keep using their existing controls, logging, and data residency setups while still enabling employees to access the new capabilities in Excel and PowerPoint.

Anthropic explicitly highlights that this gives enterprises “more flexibility” to use the add-ins within the cloud and compliance arrangements they already have in place. While the company does not go into technical detail in this announcement, the intent is clearly to fit into heterogeneous environments where different business units may be tied to different hyperscaler stacks.

For operations leaders, this model raises practical questions: which workflows should be encoded as Skills, how to govern who can create and share them, and how to evaluate the risk profile of cross-app AI actions that read live spreadsheets and modify presentations. The announcement does not answer these directly, but it does make clear that Claude is designed to live inside existing enterprise boundaries, rather than pulling work out into a separate, standalone AI workspace.

How Claude Compares to Microsoft Copilot Cowork

The timing and capabilities of Anthropic’s release inevitably invite comparison with Microsoft’s new Copilot Cowork, introduced just days earlier. Copilot Cowork is Microsoft’s agentic layer that can execute tasks across Office applications such as Excel and PowerPoint — and notably, Microsoft has said it built Copilot Cowork in conjunction with Anthropic.

Anthropic itself has also launched a standalone desktop product, Claude Cowork, for Mac and Windows. That app lets Claude access, edit, create, and move information between files on a user’s computer autonomously, under user direction.

Until now, even with autonomous agents like Claude Cowork, users often had to break work into separate steps for each application. The new shared context in the Excel and PowerPoint add-ins narrows that gap by allowing Claude to maintain a continuous session that reads live data and writes formulas across both apps simultaneously.

That said, Anthropic does not position the new Skills feature as equivalent to the “more autonomous, agentic behavior” that Microsoft is emphasizing with Copilot Cowork. Instead, Anthropic’s Office integrations emphasize structured, repeatable workflows that humans trigger and oversee from within familiar sidebars.

For organizations, the distinction is material. Copilot Cowork is framed as a more fully fledged agent operating across the Microsoft 365 estate. Anthropic’s approach, while moving in an agentic direction, remains anchored to specific, definable workflows and explicit user actions within Excel and PowerPoint — even as it leverages Anthropic models under the hood in both environments.

Anthropic’s Broader Enterprise Strategy

The upgraded Office add-ins slot into a larger pattern in Anthropic’s product roadmap. Since the launch of Claude Cowork earlier this year, the company has been making a sustained case to be a primary chat and productivity platform for enterprises.

Alongside Cowork, Anthropic offers Claude Code and the broader Claude model family. These tools have found their way into many organizations’ systems, supported by strong performance on coding benchmarks and general knowledge tasks. In practice, this means Claude can navigate a computer environment, manipulate files, and assist with complex knowledge work at scale.

Anthropic’s influence is visible even in the open-source ecosystem. OpenClaw, an open-source AI agent popular among developers, “owes much of its existence to Claude Code,” underlining how Anthropic’s tooling and models are shaping not just commercial products but also community projects.

Against this backdrop, bringing shared context and Skills into Excel and PowerPoint looks less like an isolated feature release and more like another step in a broader strategy: embedding Claude into the day-to-day tools where work actually happens, and doing so in a way that is compatible with existing enterprise infrastructure.

Anthropic is also operating in a competitive field. Google can tightly couple AI with Google Workspace (Gmail, Docs, and related tools), while Microsoft continues to lead in the Office suite and is rapidly expanding its own AI capabilities. Anthropic, which does not own a productivity suite, is instead threading itself into those environments as a layer that can be selected, configured, and governed by enterprises on their own terms.

What This Means for Enterprise AI Adoption

Stepping back, the announcement underscores a broader shift in how enterprise AI is being evaluated. The focus is no longer solely on which model tops benchmark leaderboards. Increasingly, the deciding factor is which AI systems can be trusted to get real work done across existing applications, files, and workflows.

Anthropic’s work on Claude for Excel and PowerPoint directly targets that question. Shared context reduces friction and cognitive load for users who move constantly between analysis and presentation. Skills give organizations a mechanism to capture and standardize best practices, turning idiosyncratic expertise into reusable assets. Flexible deployment via Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry allows those capabilities to plug into current governance and compliance frameworks.

For enterprise IT leaders and operations managers, the practical next steps involve experimentation and governance: piloting Skills in specific teams, mapping where shared context actually speeds up critical processes, and defining policies for how AI agents interact with sensitive spreadsheets and decks. For power users inside Microsoft Office, these updates offer a path to move from ad hoc prompting toward scalable, repeatable workflows.

As vendors like Anthropic, Microsoft, and Google continue to converge AI and productivity software, the competitive landscape is likely to hinge less on discrete features and more on how well these agents integrate with the nuanced realities of enterprise work. This release suggests Anthropic is intent on being part of that contest from inside the tools business users already rely on.