Microsoft Adds Custom Skills and Oversight to Copilot in Excel

Post LinkedIn

🖥️Read original on Computerworld

#agentic-ai#data-analysis#productivity#financial-techmicrosoft-365-copilotmicrosoftcopilotexcelonedrive

💡Learn how Microsoft is moving toward agentic AI in Excel with user-approved workflows and custom skill libraries.

⚡ 30-Second TL;DR

What Changed

Introduced 'skills' feature to automate recurring tasks like variance analysis and cash flow modeling.

Why It Matters

These features shift Copilot from a simple chatbot to a structured agentic tool, significantly reducing the risk of 'hallucinated' formulas in financial environments. It enables enterprises to standardize AI-driven reporting across teams.

What To Do Next

Create a custom SKILL.md file for your team's most repetitive Excel task to test how Copilot handles structured, domain-specific instructions.

Who should care:Enterprise & Security Teams

Key Points

•Introduced 'skills' feature to automate recurring tasks like variance analysis and cash flow modeling.
•New 'plan' mode requires user approval for AI-proposed spreadsheet changes before execution.
•Integration with third-party financial data sources including Moody’s, PitchBook, and Morningstar.
•Custom skills can be saved as SKILL.md files in OneDrive for repeatable, structured AI workflows.

🧠 Deep Insight

AI-generated analysis for this event — not the original article.

🔑 Enhanced Key Takeaways

•The 'plan' mode utilizes a multi-step reasoning architecture that separates the generation of spreadsheet formulas from their execution, effectively creating a 'human-in-the-loop' sandbox environment.
•Custom skills are powered by a new orchestration layer that allows Copilot to chain together multiple Excel functions, such as Power Query transformations and DAX measures, into a single user-defined command.
•The integration with financial data providers like Moody’s and PitchBook is facilitated through the Microsoft Graph API, enabling real-time data ingestion directly into the Excel grid without manual imports.
•The SKILL.md file format utilizes a YAML-based schema to define input parameters, expected data types, and output formatting, ensuring consistency across different user sessions.
•Microsoft has implemented a new 'Audit Trail' feature alongside these updates, which logs all AI-generated changes in a hidden sheet to assist with regulatory compliance in financial reporting.

📊 Competitor Analysis▸ Show

Feature	Microsoft Copilot (Excel)	Google Gemini (Sheets)	Tableau Pulse
Workflow Automation	Custom SKILL.md scripts	Apps Script/Gemini Extensions	Automated Insights
Human-in-the-loop	'Plan' mode (Review/Approve)	Draft/Review suggestions	Limited (Read-only)
Data Integration	Native (Moody's, PitchBook)	Google Workspace/BigQuery	Salesforce/Data Cloud
Pricing	$30/user/mo (Add-on)	Included in Gemini Business	Tiered/Enterprise

🛠️ Technical Deep Dive

Orchestration Layer: Uses a proprietary agentic framework that decomposes complex financial queries into sub-tasks mapped to specific Excel API endpoints.
Execution Sandbox: The 'plan' mode operates by generating a JSON-based transaction log that the Excel engine validates for syntax and circular references before applying changes to the workbook.
Data Connectivity: Leverages the Microsoft Graph connectors to maintain secure, authenticated sessions with third-party financial APIs, ensuring data lineage is preserved.
File Schema: SKILL.md files are parsed by the Copilot runtime as structured configuration objects, allowing for version control via standard Git workflows in OneDrive.