AI

Microsoft adds new skills — and more oversight — for Copilot in Excel

At a glance:

  • Microsoft adds a skills library and a plan‑approval mode to Copilot in Excel, targeting finance workflow automation.
  • Custom skills can be saved as SKILL.md files in OneDrive; partner skills from LSEG, Ramp and Velixo are slated for release soon.
  • The update rolls out to paid Microsoft 365 Copilot subscribers at $30 per user per month (enterprise) or $21 per user per month for organizations under 300 seats.

New skills library codifies repeatable finance processes

Microsoft unveiled a skills feature on Thursday that lets users define repeatable processes for Copilot in Excel, such as building a discounted cash flow, preparing a variance analysis, or refreshing a monthly reporting model. Brian Jones, vice president for Excel, explained that a skill guides Copilot through the required steps, applying the right structure and formatting to produce output that is easier to review, reuse, and trust. The capability moves beyond ad‑hoc prompts by turning common financial workflows into reusable assets that can be shared across teams.

Users can draw from a library of pre‑built finance skills or create their own custom skills, which are saved as SKILL.md files in OneDrive where Copilot can retrieve them. Microsoft said custom skills are already available to participants in the Insider channel and will become generally available next month. Partner vendors including LSEG, Ramp and Velixo are also building skills that Microsoft labeled as “coming soon.”

  • Building a discounted cash flow

  • Preparing a variance analysis

  • Refreshing a monthly reporting model

  • LSEG

  • Ramp

  • Velixo

Plan mode gives users oversight before AI edits spreadsheets

A new plan mode adds a governance layer by requiring Copilot to draft a list of intended actions — such as changing a formula — and then ask the user to approve, edit, or answer clarifying questions before any spreadsheet modification occurs. Jones noted that this step‑by‑step preview is designed to increase confidence in AI‑driven changes, especially when dealing with sensitive financial data. The assistant presents the plan in the chat pane, allowing the user to intervene at any point before execution.

Once the user approves the plan, Copilot carries out the actions and then posts a link to the changes in the chat window, making the audit trail explicit. All edits made by the AI appear alongside human edits in the existing Show Changes pane, so reviewers can see a unified history of modifications. This integration with Excel’s native change‑tracking reinforces compliance and simplifies post‑hoc reviews.

  • Changing a formula

Third‑party data connectors expand Copilot’s analytical reach

Copilot can now pull live data from external providers, a move that positions the assistant as a more comprehensive research tool for analysts. Supported sources at launch include Moody’s, CB Insights, Morningstar and PitchBook, each offering specialized financial, market‑intelligence and venture‑capital datasets. Users can invoke these connectors directly from the Copilot chat, blending external figures with internal spreadsheet models without leaving Excel.

The integration eliminates the manual copy‑paste cycle that often introduces errors and version‑control issues. By keeping the data lineage visible in the chat history, teams can trace how external inputs influenced a model’s outcome. Microsoft indicated that additional connectors are on the roadmap, suggesting a broader ecosystem play.

  • Moody’s
  • CB Insights
  • Morningstar
  • PitchBook

Rollout timeline pricing and availability

The new skills, plan mode and data connectors will roll out progressively to customers who hold a paid Microsoft 365 Copilot license. Microsoft offers two subscription tiers: $30 per user per month for larger enterprises and a Microsoft 365 Copilot Business plan at $21 per user per month for organizations with fewer than 300 employees. Both tiers grant access to the full suite of Copilot capabilities in Excel, including the newly announced features.

Microsoft 365 Copilot became generally available in Excel in late 2024, and since then the company has layered on agentic tools, a native Copilot function and Python support for advanced data analysis. The latest update builds on that foundation by adding governance and domain‑specific automation, addressing feedback from finance teams that demanded more control and repeatability. Analysts expect the enhanced oversight and skill framework to accelerate adoption in regulated industries where audit trails are mandatory.

  • $30 per user per month (enterprise)
  • $21 per user per month (Business, <300 employees)
Editorial SiliconFeed is an automated feed: facts are checked against sources; copy is normalized and lightly edited for readers.

FAQ

What new skills does Microsoft 365 Copilot in Excel provide for finance teams?
Microsoft introduced a skills library that lets users codify repeatable finance processes such as building a discounted cash flow, preparing a variance analysis, or refreshing a monthly reporting model; custom skills can be saved as SKILL.md files in OneDrive and partner skills from LSEG, Ramp and Velixo are slated for release soon.
How does the new plan mode increase oversight of Copilot actions in Excel?
Plan mode forces Copilot to draft a step‑by‑step list of intended spreadsheet changes — like modifying a formula — and then prompts the user to approve, edit, or answer clarifying questions before any edit is applied; after execution, a link to the changes appears in the chat and all AI edits are shown alongside human edits in the Show Changes pane.
What are the pricing options for Microsoft 365 Copilot after this update?
Paid Microsoft 365 Copilot subscribers can choose between a $30 per user per month enterprise tier and a $21 per user per month Business tier for organizations with fewer than 300 employees; both tiers include access to the new skills, plan mode and third‑party data connectors.

More in the feed

Prepared by the editorial stack from public data and external sources.

Original article