Analyst Agent in M365 Copilot: Quick Start Guide

At a glance

What it is
A data analysis feature. Upload a file,
ask a question, receive charts, tables, and calculation results.

Key difference from Chat
Chat summarises M365 data you already have.
Analyst performs calculations using Microsoft-managed analytical processing on uploaded files.

Key difference from Excel Copilot
Excel Copilot is available within an open Excel workbook.
Analyst supports uploaded files within the Copilot chat interface.

Use case	What Analyst does	For
Cost or budget variance	Identifies over/under-budget lines, calculates variances	Finance · PM · Commercial
Trend analysis	Finds patterns across time series data, highlights anomalies	Analyst · Finance · Operations
Forecast	Projects values forward based on historical data	Finance · Sales · Planning
Data comparison	Cross-references two datasets, finds differences	Commercial · Procurement · HR
Report summary	Pulls key metrics from a PDF or PowerPoint	Executive · Manager · EA

What's new — June 2026. Copilot Cowork is now generally available and billed on usage (Copilot Credits, $0.01/credit pay-as-you-go), and the Work IQ APIs reached GA. If you build custom agents that ground in your M365 data via Work IQ, that usage is metered in Copilot Credits. Analyst itself is a pre-built Microsoft agent under your Copilot licence and is not separately metered — but the new usage billing applies tenant-wide, so for agent billing and credit consumption explained, see the cost model reference.

Analyst vs Chat vs Excel Copilot — when to use each

Tool	Best for	Input	Output
Analyst	Deep data analysis — calculation, trends, charts — on any uploaded file	Uploaded files (.xlsx, .csv, .pdf, .pptx, .xml)	Charts, tables, insights, forecasts
Chat	Quick answers from your live M365 data — email, calendar, Teams	Your live M365 data	Conversational — no file upload
Excel Copilot	In-workbook analysis — formulas, highlighted cells, pivot tables	The open Excel workbook	Formulas, formatting, pivot tables — inside Excel

How to use Analyst

Open Copilot and select Analyst

Go to m365.cloud.microsoft. In the left panel under Agents, select Analyst. If you do not see it, click All agents — it is built-in and does not require manual installation.

Upload your file

Click the + button in the chat input. It offers three routes: Add work content (files, emails, and meetings from your tenant), Upload images and files (from your device), or Attach cloud files (OneDrive). Upload the file you want to analyse: Excel workbook (.xlsx), CSV, PDF, PowerPoint, or XML. You can upload multiple files in one session. Note: device uploads are copied to your OneDrive (work/school), and Copilot shows a banner saying so at upload time.

Ask your question

Type your question directly in the chat. Be specific: "Show me total spend by category" works better than "analyse this". To see the code behind an answer, expand the reasoning panel ("Reasoning completed in N steps") above the response; the executed Python sits under the Coding and executing step, with a copy button. (In the previous design this was a standalone "Show code" button on the response.) Analyst uses Microsoft-managed code execution in a sandboxed environment — external packages and internet access are not available within the execution context.

Iterate

Ask follow-up questions in the same session. Analyst remembers the context and the uploaded files. You can ask it to filter, recalculate, change the chart type, or break down a specific category further.

Export results

Copy charts or tables from the chat window, or open the response's ⋯ menu: Edit in Pages turns the output into a shareable Copilot Page, Export to sends it to Word, and Schedule this prompt re-runs the request on a timer. You can also copy individual outputs into PowerPoint or email.

Supported file types

.xlsx / .csv Best results — structured data with column headers. Large files supported.

.pdf Tables in PDFs are readable. Scanned PDFs (image-only) may not parse correctly.

.pptx Extracts text and table data from slides. Charts embedded as images are not readable.

.xml Structured XML data. Works well for export files from project systems (P6, SAP).

.docx Text and tables from Word documents. Useful for summarising report data.

Multiple files Upload two or more files in one session to cross-reference datasets.

Use cases — what to upload and what to ask

Each card shows example files and example prompts

Use cases — by role and scenario

Budget and cost variance

Finance · PM · Commercial Manager

Most common use case

Upload these

SAP cost export (CSV or Excel)
Approved budget spreadsheet
Forecast update (latest period)

Ask this

Which cost codes are over budget by more than 10%?
Show me total spend by category as a bar chart.
Compare actual vs forecast for each work package and flag variances above threshold.
What is the total committed cost excluding contingency?

Clean column headers make a significant difference. Rename columns before uploading if your SAP export uses internal codes — Analyst interprets column names as context.

Sales pipeline and performance

Sales · BD · Revenue Operations

High value for deal tracking

Upload these

CRM export (CSV) with deal stage and value
Monthly close data for the last 12 months
Rep performance spreadsheet

Ask this

Show me pipeline by stage and expected close month.
Which deals have been in the same stage for more than 30 days?
What is the win rate by deal size for the last 6 months?
Forecast Q3 revenue based on current pipeline and historical close rates.

Analyst does not connect live to Salesforce, HubSpot, or Dynamics — export your data to CSV first. For real-time CRM analysis, use those platforms' native AI tools.

Project schedule and progress

PM · Programme Director · PMO

Useful for P6 / Primavera exports

Upload these

Primavera P6 schedule export (XER → CSV or Excel)
Progress update spreadsheet (% complete by activity)
Previous period baseline data

Ask this

How many activities are behind schedule and by how much?
Show me critical path activities with negative float.
What is the schedule performance index by work area?
Which activities have slipped since the last period update?

Export from P6 as CSV or Excel with readable column names. Very large schedules (10,000+ activities) may need to be filtered before uploading — Analyst works best on focused datasets.

Known limitations and considerations

The output looks wrong or the numbers do not match. Analyst generates Python code to process your data. Expand the reasoning panel and open the Coding and executing step to inspect what it calculated (in the previous design, a standalone "Show code" button). If the code uses a column name incorrectly (e.g., summing the wrong field), ask it to recalculate with the correct column specified explicitly.

A scanned PDF does not parse. PDFs that are image-based (scanned documents without OCR text layer) cannot be read by Analyst. Convert to a text-based PDF using your scanner software's OCR feature, or type the data into a CSV instead.

Charts look different from what you expected. Analyst chooses chart types automatically. If the default is not right, ask explicitly: "Show me this as a line chart with month on the x-axis" or "Use a stacked bar chart grouped by category."

Large files slow the session or hit a limit. Very large files (50MB+, or spreadsheets with hundreds of thousands of rows) may time out or return incomplete results. Filter or aggregate your data before uploading — upload the slice you need, not the entire dataset.

The session forgets your earlier uploads. Analyst retains uploaded files within one conversation session. If you start a new conversation, you need to re-upload your files. Do not close the tab mid-session if you plan to continue iterating.