Microsoft is doubling down on AI-driven productivity with a sweeping update to Copilot in Excel, announced on June 25, 2026. The new capabilities—reusable Skills, finance-oriented data connectors, planning controls, and change attribution—will roll out to Excel for the web and Windows, bringing advanced automation to the world's most ubiquitous spreadsheet application. These enhancements are not mere tweaks; they represent a fundamental shift in how finance professionals, data analysts, and business users will interact with Excel, moving from manual formula crafting to conversational, agentic workflows.

Reusable Skills: Automating Repetitive Tasks with AI

Excel has long supported automation through macros and VBA, but Copilot Skills take the concept into the AI era. A Skill is a user-defined set of instructions, written in natural language, that Copilot can execute on demand across any workbook. Microsoft describes it as a macro that you teach by conversation. Users can create a Skill by describing what they want in plain English, for example: \"Clean this dataset by removing duplicates, standardize date formats to YYYY-MM-DD, and highlight cells with values above the 90th percentile.\" Once saved, that Skill can be invoked in other workbooks, even if the data structure differs, because Copilot intelligently maps the intent to the actual columns and ranges.

In Excel for the web, a new 'Skills' pane appears alongside Copilot. Users can record a Skill by demonstrating steps, or they can describe a sequence in natural language. Copilot then generates a Skill definition card with a name, description, and example invocation. Users test the Skill on sample data before saving it to their personal library or sharing it with their team via SharePoint or OneDrive. Organizations can build libraries of tailored Skills for consistent data processing across departments, potentially saving thousands of hours currently spent on monotonous data preparation.

Moreover, Skills are not static—they can be refined over time. If a particular step doesn't work as expected, the user tweaks the Skill without starting over. Microsoft envisions a future where a marketplace of verified Skills exists, similar to the Office Add-in store, further expanding Copilot's utility.

Finance Data Connectors: Bridging the Gap Between Excel and Enterprise Data

The new finance-oriented data connectors mark a strategic push into the CFO's office. Copilot can now connect directly to external financial data platforms such as Bloomberg Terminal, S&P Capital IQ, and internal ERP systems like SAP and Oracle. With a simple prompt—\"pull the latest balance sheet for MSFT, AAPL, and GOOGL into a new sheet and compare their debt-to-equity ratios\"—Copilot handles authentication, data retrieval, and formatting.

What sets these connectors apart is their contextual awareness. Copilot monitors your analysis and suggests relevant data. If you're building a discounted cash flow model, it might offer to import risk-free rates, industry betas, and peer multiples. Administrators enable connectors via the Microsoft 365 admin center, configuring authentication parameters. Users access them through the Data tab or directly in the Copilot chat by referencing the connector name.

Microsoft has emphasized security: all connections adhere to Microsoft 365's data loss prevention policies, and sensitive credentials are stored in Azure Key Vault. Data is encrypted in transit and at rest, and connectors comply with GDPR and other regulations. This allays fears of financial data leakage while providing seamless integration that could replace manual CSV imports and legacy ODBC connections.

Planning Controls: Turning Excel into a Scenario Simulation Engine

Financial planning and analysis (FP&A) professionals have long used Excel for scenario modeling, but the process is often manual and error-prone. The new planning controls let Copilot manage multiple \"what-if\" parameters simultaneously. In the Copilot sidebar, a 'Scenario' tab allows users to define variables, assign data ranges, and set comparison criteria. Copilot then generates a dynamic dashboard with sensitivity tables and waterfall charts.

For instance, an FP&A lead could say: \"Create three scenarios for next year's budget—pessimistic, base, and optimistic—with revenue growth at 2%, 5%, and 8% respectively, and fixed cost inflation at 3%. Show me the impact on EBITDA and free cash flow.\" Copilot will not only compute the numbers but also annotate the results with trade-offs and risks. This goes beyond Excel's built-in Scenario Manager by integrating AI-driven insights and automatically updating linked sheets.

Planning controls support iterative refinement—users can adjust assumptions on the fly and see real-time recalculations. The feature also explains its reasoning in plain English, making it easier for non-finance stakeholders to grasp complex models. By reducing the grunt work of scenario building, finance teams can spend more time interpreting results and crafting strategies.

Change Attribution: Building Trust Through Transparency

One of the biggest barriers to AI adoption in finance is the \"black box\" problem. Executives need to know why a number changed and who or what caused it. Change attribution directly tackles this issue. Every modification made by Copilot—whether it's a generated formula, a data refresh, or a restructured table—is logged with a detailed trail. In the review tab, a new 'Changes' dropdown lists all AI-driven modifications with diff views. Hovering over a cell reveals a card showing the original value, the new value, the timestamp, the user who invoked Copilot, and the prompt or Skill that initiated the change.

This functionality is analogous to collaborative editing features but specifically tuned for AI contributions. Audit-ready industries like banking and insurance will find this indispensable. Furthermore, change attribution integrates with Microsoft Purview compliance tools, enabling eDiscovery and retention policies for AI-generated content. Each change includes a trust score indicating the model's confidence, although the user is always the final arbiter. Early adopters have praised this as a \"must-have\" for regulated environments.

Platforms and Rollout Timeline

The features will be available first in Excel for the web and on Windows desktop, with Mac and mobile versions to follow. Microsoft has not provided a specific timeline for broader platform support, but given the company's recent emphasis on cross-platform consistency, it's likely to arrive within the year. Licensing details remain under wraps; however, full Copilot functionality typically requires a Microsoft 365 Copilot license, which costs $30 per user per month for enterprise customers. It's unclear whether certain Skills or connectors will be gated behind premium tiers.

The Agentic Office Vision

These updates are part of what Microsoft CEO Satya Nadella has called the \"agentic Office\"—a suite where AI agents proactively assist users. Copilot in Excel is no longer just a Q&A bot; it can now perform multi-step workflows, reason over data, and explain its actions. Combined with Copilot in Word, PowerPoint, and Teams, the ecosystem is moving toward autonomous knowledge work. Analysts predict that by 2028, 40% of spreadsheet tasks will be initiated by AI rather than humans.

User Reactions and Potential Challenges

While the announcement has generated considerable excitement, especially among Excel power users, some pragmatic concerns have emerged. On Hacker News and Reddit, users questioned the reliability of AI-generated scenarios in high-stakes financial modeling. \"What if Copilot hallucinates a formula that looks right but is subtly wrong?\" one commenter wrote. Change attribution partially mitigates this risk by making errors traceable, but it doesn't eliminate the need for human oversight.

Another common worry is the learning curve. Although Skills are designed to be intuitive, creating effective prompts remains a skill in itself. Microsoft plans to offer templates and a \"Skill gallery\" to jumpstart adoption, but organizations will need to invest in training. Data security with third-party connectors also drew scrutiny; Microsoft asserts that data is encrypted and compliant, but IT departments will want to thoroughly vet each connection.

Competitive Context

Microsoft is not alone in the AI spreadsheet race. Google Sheets has been infusing its Gemini AI with natural language formulas and smart fill. Apple's Numbers added limited AI suggestions last year. However, Excel's unmatched user base and deep enterprise integration give Copilot a significant advantage. The new Skills and connectors widen the gap, especially for finance professionals who rely on Excel's advanced data modeling capabilities.

What This Means for Excel's Future

These features mark Excel's transition from a passive calculation engine to an active analytical partner. For decades, Excel users have manually wrestled with data—now, they can delegate routine tasks to Copilot and focus on strategic decisions. The reusable Skills concept has the potential to democratize automation, enabling even non-technical users to build custom AI assistants. As Microsoft refines the underlying language models, expect Skills to become more context-aware and adaptive.

In the near term, the biggest impact will be felt in finance departments, where the combination of live data connectors, planning controls, and transparent attribution addresses core workflow pain points. Longer term, these capabilities could reshape how business students and professionals are trained on Excel, shifting the curriculum from formula syntax to prompt engineering and model validation.

Conclusion

Microsoft's June 25, 2026, announcement of Copilot Skills, finance data connectors, planning controls, and change attribution for Excel is a bold step toward an AI-first productivity suite. By embedding these features into Excel for the web and Windows, Microsoft is meeting finance professionals where they already work—and giving them tools that promise to reduce grunt work, increase accuracy, and improve decision-making. The proof, as always, will be in the execution: how reliable Copilot proves in messy real-world data, and how quickly users adopt a conversational approach to the spreadsheets they've managed manually for so long.