Microsoft has begun rolling out a major update to Copilot in Excel that adds finance-specific Skills, a new Plans feature for structured modeling, and direct connections to external data sources—all with an eye toward auditability and enterprise compliance. The enhancements, now available to paid Microsoft 365 Copilot subscribers, mark a decisive pivot from generic AI assistance to domain-specific, governance-first tooling for finance professionals.
Custom Skills—reusable prompt templates tailored to individual business needs—are already live in Insider builds, giving early adopters a head start in automating complex financial workflows. The move underscores Microsoft’s ambition to make Excel not just a canvas for numbers but an intelligent, auditable platform for critical business decisions.
What’s New in Excel Copilot
The latest wave of Copilot features targets the finance department’s most pressing needs: accuracy, repeatability, and traceability. At its core are three pillars: Finance Skills, Plans, and external data connectivity. Together, they transform Excel from a static spreadsheet tool into a dynamic, AI-driven environment where models are built with guardrails, and every action is logged for review.
Microsoft first introduced Copilot in Excel in March 2024 with natural language formula generation and basic data analysis. The initial release, while powerful, lacked the depth required by enterprise finance teams—users couldn’t easily reuse complex prompts, connect to live data sources, or prove that outputs were reliable. The new features directly address those gaps.
Finance Skills: Purpose-Built for the Back Office
Finance Skills are pre-configured AI capabilities designed for common financial tasks. Rather than starting from a blank prompt bar, users can select from a library of skills such as variance analysis, cash flow forecasting, depreciation modeling, and what-if scenario planning. Each skill encapsulates best practices—for example, a forecasting skill might automatically apply seasonality adjustments and Monte Carlo simulations, with the underlying methodology visible to the user.
These skills are not black boxes. Microsoft has built in transparency features that show which formulas and data ranges were used, making it possible for a colleague or auditor to pick apart the model. In practice, a controller can ask Copilot to “create a rolling 12-month revenue forecast based on historical data and current pipeline,” and the skill will generate the output while exposing the step-by-step logic.
Plans: Structured Modeling with an Audit Trail
Perhaps the most significant addition is Plans—a new approach to building, sharing, and governing financial models. A Plan is a container for a set of related Copilot interactions, data connections, and manual inputs, all organized into a repeatable workflow. Think of it as a smart template that remembers how you arrived at a conclusion.
When a finance team creates a quarterly earnings model, they can now package it as a Plan. The Plan records every Copilot query, data refresh, and assumption change. Version history is maintained automatically, and collaborators can add comments or approvals directly within the Plan interface. For internal audit and SEC compliance, this traceability is invaluable. The Plan can be locked, exported as a PDF with a full audit log, or integrated into a broader Power Automate approval flow.
Microsoft has also introduced the concept of “visible Plans,” meaning that the logic and flow of a model are graphically represented. Users can see a flowchart of how data moved through various Copilot steps, which skills were applied, and where human overrides occurred. This visual audit trail lowers the barrier for non-technical stakeholders to understand complex models.
External Data Connections: Breaking Down Silos
Finance models are only as good as the data feeding them. Recognizing that critical data often lives outside Excel, Microsoft is enabling direct connections to external sources via Copilot. Users can now link to data warehouses, ERP systems like Dynamics 365 and SAP, and even third-party APIs—all without writing SQL or VBA.
A procurement analyst could, for instance, ask Copilot to “pull the latest vendor spending from SAP by category and compare it against budget thresholds.” The connection is established securely through Microsoft Entra ID, and data refreshes can be scheduled or triggered on demand. Importantly, the lineage of imported data is maintained within the Plan, so any subsequent audit can trace a number back to its original source system and timestamp.
Built for Auditability: The Compliance Backbone
The phrase “built for auditability” is not marketing fluff. Every Copilot action in Excel now generates metadata that can be consumed by compliance tools. From the moment a skill is invoked to the final cell value, the system logs:
- The prompt text and parameters
- The AI model version used
- The snapshot of input data at the time of execution
- Intermediate calculations and the final output
- User identity and timestamp
These logs are stored in accordance with the Microsoft 365 Compliance Center, meaning they are discoverable via eDiscovery and can be exported for external auditors. For finance teams subject to SOX, GDPR, or other frameworks, this transforms Copilot from a potential liability into a compliant, auditable resource.
Custom Skills: Tailoring Copilot to Your Business
For organizations with unique processes, Microsoft has added the ability to create custom Skills. Insiders can now define and publish their own skill packages using a low-code builder. A healthcare provider might build a skill for calculating patient revenue cycles, while a manufacturing firm could create one for materials requirement planning.
Custom Skills can be centrally managed via the Microsoft 365 Admin Center and deployed to all users or specific groups. This ensures that the AI behaves consistently across the organization, using approved formulas and data connections. Microsoft has hinted that a marketplace for third-party Skills is on the horizon, which could open the door to specialized skills from industry consultants and ISVs.
Availability and Licensing
All new capabilities are included with a Microsoft 365 Copilot license ($30 per user per month as an add-on to existing Microsoft 365 E3/E5 plans). The Finance Skills and Plans features are currently rolling out to general availability, while custom Skills remain in preview for Windows, Mac, and web users enrolled in the Microsoft 365 Insider program. Microsoft has not announced specific dates for broad custom Skills availability but suggests it will follow by the end of the quarter.
To use the external data connection features, organizations must have the appropriate data source licensing (e.g., SAP, Dynamics 365) and configure connectors through Power Platform. The connections comply with existing data loss prevention policies, so security-conscious enterprises can define which sources are permitted.
The Bigger Picture for Enterprise AI
These updates position Excel Copilot as more than just a productivity booster—it becomes a strategic platform for AI-augmented financial planning and analysis. By embedding auditability and repeatability into the core design, Microsoft is directly addressing the barriers that have kept regulated industries from embracing generative AI.
Competitors like Google Sheets with Gemini and specialized tools like Anaplan or Pigment have long touted collaboration and planning features, but Excel’s installed base of over a billion users gives Copilot a distribution advantage. The addition of Plans and external data connections narrows the gap with dedicated FP&A platforms, while the audit capabilities may tip the scales for risk-averse CFOs.
What This Means for Finance Teams
For the professionals in the trenches, these features could fundamentally change daily workflows. Instead of spending hours manually updating spreadsheets, finance analysts will be able to delegate data collection, transformation, and initial analysis to Copilot. The time saved can be redirected toward strategic interpretation and decision-making.
However, success will depend on adoption and cultural change. Microsoft is bundling Copilot with existing subscriptions, which lowers the barrier, but training users to trust and verify AI outputs remains a challenge. The visible Plan structure and audit logs are designed to build that trust over time. Finance leaders who embrace these tools early may gain a competitive edge in speed and accuracy.
Looking ahead, Microsoft has signaled that Copilot’s finance capabilities will expand to include deeper integration with Power BI for dashboarding and with Teams for collaborative review sessions. The trajectory is clear: Excel is evolving from a standalone application to the front door of a connected, AI-first finance ecosystem.