← Back to Index

Copilot in Excel for Finance

Hands-On Training Labs — From Data Prep to Month-End Close

Scenario: You're a finance professional at Contoso Corp, a mid-market company with ~$180M annual revenue. It's the first week of April 2026 and your team is closing March / Q3 FY26 books. You'll use Copilot in Excel to clean GL data, analyze budget variances, age receivables, generate financial commentary, and reconcile the bank statement — the same workflows you do every month, but faster.
75–90 minutes Copilot in Excel (Edit with Copilot + COPILOT() Function) Instructor-Led or Self-Paced

Setup
1

GL Cleanup
2

Variance
3

AR Aging
4

COPILOT()
5

Bank Rec

How to Use This Guide

Each lab is collapsible — click the title to expand or collapse. Every prompt has a Copy button so you can paste it straight into Excel Copilot. Work through Lab 0 first for setup, then complete the labs in order — they build on each other using the same company scenario.

💡
Pro tip: The most productive Copilot users treat it as a conversation. Start broad, refine based on results, and iterate until the output matches what you'd present to leadership. If Copilot doesn't nail it on the first try, rephrase — just like you would with a colleague.

Learning Objectives

By the end of this training, you will be able to:

Lab 0 — Quick Start & Setup

What is Edit with Copilot?

⏱️ 5 minutes 🎯 Setup & orientation 🛠️ No exercise — just read

Edit with Copilot is an AI assistant built into Excel. Instead of writing formulas or building charts yourself, you describe what you want in plain English — and Copilot makes the change directly on the sheet. You stay in control: every change can be reviewed, kept, or undone with Ctrl+Z.

Key idea: Copilot doesn't just tell you what to do — it does it for you. And every formula it writes is a real Excel formula you can inspect and edit.

Before you start — check these

🔑Microsoft 365 Copilot license
📊Excel for the web, Windows, or Mac
☁️Files saved to OneDrive or SharePoint
🌐Internet connectivity
📋Data formatted as an Excel Table
💡
Tip: Press Ctrl+T to convert a data range into a table. Copilot performs significantly better with tables — it understands column names and data boundaries automatically.

How to open Copilot in Excel

  1. Open your workbook in Excel.
  2. Click the Copilot icon on the Home tab of the ribbon.
  3. The Copilot pane opens on the right side of the screen.
  4. Type a natural-language prompt describing what you want. Copilot processes it and makes changes on the sheet.
  5. If you don't like the result, press Ctrl+Z to undo instantly.
📌
Don't see the Copilot icon? Your tenant or license may not have Copilot enabled — check with your IT admin. Also ensure the file is saved to OneDrive or SharePoint (not stored locally).
🔴
Important: Lab 4 uses the new COPILOT() function, which is currently in Preview. You'll need to be on the right Microsoft 365 channel (Beta / Current Channel Preview) for that lab to work. Labs 0–3 and 5 work in standard channels.

Download the Sample Workbooks

Download the two Excel workbooks used in this training. Save them to your OneDrive before opening.

📥 Finance-Copilot-Labs-1-3.xlsx 📥 Finance-Copilot-Labs-4-6.xlsx
📌
Important: After downloading, upload each file to your OneDrive and open from there (or open directly in Excel for the web). Copilot requires the file to be stored in OneDrive or SharePoint — it won't work with local-only files.

Instructor Note

Before the session: verify both files open correctly, confirm Copilot is available in the Excel ribbon, and pre-load the first workbook to minimize setup time. Walk through the Copilot pane orientation (prompt box, suggestions, Apply/Insert Column buttons) before Lab 1.

Lab 1 — Clean & Structure Raw Financial Data

Clean & Structure a Raw GL Export

⏱️ 15 minutes 🎯 Data transformation & enrichment 📄 Sheet: GLExport

Why This Matters

Every finance team's month starts the same way: a raw data pull from the ERP that needs cleanup before anyone can analyze it. Dates are inconsistent, account codes need mapping to reporting categories, and descriptions are cluttered. This is the kind of repetitive prep work where Copilot saves hours — and where most people don't realize AI can help.

Time comparison

⏳ Manual: 30–45 minutes of formula writing and data manipulation ⚡ With Copilot: 5–8 minutes of prompting and reviewing

Instructor Talking Point

Emphasize that Copilot writes real Excel formulas. After each exercise, have participants click into the new column and check the formula bar. They'll see functions like YEAR(), IF(), and TEXTAFTER() — formulas they can inspect, edit, and learn from.

Open the File

  1. Open Finance-Copilot-Labs-1-3.xlsx from your OneDrive.
  2. Navigate to the "GLExport" sheet tab.
  3. Click the Copilot button in the ribbon to open the Copilot pane.
  4. Take a moment to scan the data — notice the inconsistent date formats, raw account codes, and verbose descriptions. This is typical of what comes out of an ERP.

1.1 Add Fiscal Year and Fiscal Quarter

Your company's fiscal year starts July 1. You need fiscal period columns for reporting.

📋 Prompt Add a column called "Fiscal Year" based on the Date column. Our fiscal year starts in July — for example, July 2025 through June 2026 is FY2026.

→ Click "Insert Column" to add it.

Now add the quarter:

📋 Prompt Add a column called "Fiscal Quarter" using the Date column. Our fiscal year starts in July, so Jul-Sep is Q1, Oct-Dec is Q2, Jan-Mar is Q3, Apr-Jun is Q4.

→ Click "Insert Column"

What to expect

  • Two new columns: "Fiscal Year" (e.g., FY2026) and "Fiscal Quarter" (e.g., Q3)
  • Each cell contains a formula — click one to see the logic in the formula bar
  • Jan–Mar 2026 entries should show FY2026 / Q3

1.2 Map Account Codes to Reporting Categories

Leadership doesn't read 4-digit account codes — they need categories like "Revenue" and "Operating Expenses."

📋 Prompt Add a column called "Reporting Category" that groups the Account Code into these categories: - 4000–4999 = "Revenue" - 5000–5999 = "Cost of Goods Sold" - 6000–6999 = "Operating Expenses" - 7000–7999 = "Other Income/Expense"

→ Click "Insert Column"

💡
Tip: Check the formula Copilot generates. It will likely use nested IF() or IFS() — formulas that would take most people several minutes to write and test manually. This is a great example of Copilot handling tedious-but-necessary logic.

1.3 Extract Vendor Names from Descriptions

The Description field contains valuable information buried in free text. Let's pull out the vendor or counterparty name.

📋 Prompt Add a column called "Vendor/Counterparty" that extracts the company or person name from the Description column. Look for names after patterns like "Invoice" references, "PO-" references, dashes, or at the end of the description.

→ Click "Insert Column"

What to expect

  • A new column with extracted names like "Acme Corp", "J.Smith", etc.
  • Some cells may be blank or imperfect — that's normal with unstructured text
  • This demonstrates Copilot's ability to parse free-text fields
📌
Real-world note: Text extraction from messy descriptions is one of those tasks where Copilot won't be 100% accurate — but getting 80% right in seconds beats manually reviewing every row. Always spot-check AI-extracted data.

1.4 Flag Large Transactions for Review

Your team's policy requires review of any journal entry over $10,000. Let's flag them automatically.

📋 Prompt Add a column called "Review Flag" that shows "REVIEW" if the Debit or Credit amount is greater than or equal to $10,000, and blank otherwise.

→ Click "Insert Column"

Now make it visual:

📋 Prompt Highlight all rows where the Review Flag column says "REVIEW" in light red.

Self-Check — Lab 1

Lab 2 — Budget vs. Actuals Variance Analysis

Analyze Budget Variances Like a Pro

⏱️ 15 minutes 🎯 Variance analysis & summarization 📄 Sheet: BudgetVsActuals

Why This Matters

Budget-vs-actuals variance analysis is the single most common finance reporting workflow. Every department head, every controller, every CFO sees a version of this report every month. The mechanics — calculating variances, flagging outliers, building summaries — are straightforward but time-consuming. Copilot handles the mechanics so you can focus on the story the numbers tell.

Time comparison

⏳ Manual: 45–60 minutes for formulas, formatting, summary, and chart ⚡ With Copilot: 10–15 minutes of prompting, reviewing, and refining

Open the Sheet

  1. In the same workbook (Finance-Copilot-Labs-1-3.xlsx), switch to the "BudgetVsActuals" sheet tab.
  2. Review the data: departments, accounts, and monthly budget/actual figures for January through March.

2.1 Calculate Q1 Variance Dollars and Percentages

First, let's add the core variance calculations.

📋 Prompt Add a column called "Q1 Variance $" that calculates Q1 Actual minus Q1 Budget. Then add another column called "Q1 Variance %" that calculates the variance as a percentage of budget.

→ Click "Insert Column" for each.

What to expect

  • Two new columns with variance calculations
  • Negative values indicate under-budget (favorable for expenses, unfavorable for revenue)
  • Percentages formatted as decimals (you may want to format as %)

2.2 Create a Department Summary

Roll up the detail into a leadership-ready department summary.

📋 Prompt Create a summary table showing total Q1 Budget, total Q1 Actual, total Q1 Variance $, and Q1 Variance % grouped by Department.

→ Click "Add to a new Data Sheet"

💡
Tip: Navigate to the new summary sheet to verify the data. Then come back to the BudgetVsActuals sheet for the next exercise.

2.3 Identify the Top Unfavorable Variances

The CFO always asks: "What's driving the miss?" Let's answer that proactively.

📋 Prompt Which 5 line items have the largest unfavorable Q1 variance in absolute dollar terms? Show the department, account name, Q1 Budget, Q1 Actual, and variance for each.

What to expect

  • Copilot identifies the top 5 problem areas with specific dollar figures
  • For expense accounts, unfavorable = over budget (actual > budget)
  • For revenue accounts, unfavorable = under plan (actual < budget)

2.4 Flag Items Exceeding the 10% Threshold

Your finance policy requires explanation for any line item with variance exceeding ±10% of budget.

📋 Prompt Highlight all rows where the absolute Q1 Variance % exceeds 10% in red.

What to expect

  • Rows with variances beyond ±10% highlighted in red
  • These are the items that need written explanations in the variance report

2.5 Visualize the Variance Story

A chart makes the variance story immediately clear in presentations.

📋 Prompt Create a bar chart showing Q1 Variance $ by Department. Include both positive and negative variances so I can see which departments are over and under budget.

What to expect

  • A bar chart with departments on the axis and variance dollars as bars
  • Over-budget departments show as positive (or red) bars; under-budget as negative (or green)
  • Immediately tells the leadership story: who's on track, who isn't

Instructor Talking Point

This is a good moment to pause and ask participants: "How long would this entire sequence — variance formulas, department summary, top 5 analysis, conditional formatting, and chart — take you manually?" Most finance teams say 45–60 minutes. With Copilot, the class just did it in under 10.

Self-Check — Lab 2

Lab 3 — Accounts Receivable Aging & Collections

Build an AR Aging Report with Collection Priorities

⏱️ 15 minutes 🎯 Date math, categorization & analysis 📄 Sheet: ARDetail

Why This Matters

Cash is oxygen. AR aging reports drive collections strategy, reserve calculations, and cash flow forecasting. Building one requires date math, bucket logic, and prioritization — exactly the kind of multi-step work where Copilot shines. This lab takes you from raw invoice data to a collections-ready report with priorities.

Time comparison

⏳ Manual: 30–40 minutes of DATEDIF formulas, nested IFs, and pivot tables ⚡ With Copilot: 8–12 minutes of guided prompting

Open the Sheet

  1. Stay in Finance-Copilot-Labs-1-3.xlsx and switch to the "ARDetail" sheet tab.
  2. Review the invoice data — note which invoices are paid, unpaid, and partially paid.

3.1 Calculate Outstanding Balance and Days Outstanding

First, determine how much is still owed and how overdue each invoice is.

📋 Prompt Add a column called "Outstanding Balance" that calculates Invoice Amount minus Amount Paid. If Amount Paid is blank, use the full Invoice Amount.

→ Click "Insert Column"

📋 Prompt Add a column called "Days Outstanding" that calculates the number of days between the Due Date and today's date. If the invoice has been fully paid, show 0.

→ Click "Insert Column"

3.2 Create Aging Buckets

Standard aging buckets are the foundation of any AR report.

📋 Prompt Add a column called "Aging Bucket" that categorizes each invoice based on Days Outstanding: - 0 or negative = "Current" - 1–30 = "1-30 Days" - 31–60 = "31-60 Days" - 61–90 = "61-90 Days" - Over 90 = "90+ Days" If the invoice is fully paid, show "Paid".

→ Click "Insert Column"

What to expect

  • A new column with standard aging bucket labels
  • Paid invoices show "Paid" regardless of dates
  • The formula likely uses nested IF() or IFS() — check the formula bar

3.3 Summarize Total Exposure by Aging Bucket

📋 Prompt Create a summary table showing total Outstanding Balance by Aging Bucket. Include the count of invoices in each bucket. Sort from most overdue to current.

→ Click "Add to a new Data Sheet"

3.4 Identify Top Collections Priorities

The collections team needs to know where to focus. Large, old invoices get priority.

📋 Prompt Which unpaid invoices have the highest outstanding balance and are more than 60 days past due? Show the customer name, invoice number, outstanding balance, and days outstanding. Sort by outstanding balance descending.

What to expect

  • A prioritized list of the riskiest receivables
  • Focus on large balances in the 61-90 and 90+ day buckets
  • This is exactly what a collections manager needs for their Monday morning call list

3.5 Visualize the Aging Distribution

📋 Prompt Create a chart showing outstanding balance by aging bucket. Use a bar or column chart to show the concentration of receivables by age.
💡
Tip: This chart is a staple of every monthly finance package. Combined with the summary table from Exercise 3.3, you now have a complete AR aging report — built in minutes, not hours.

Self-Check — Lab 3

Lab 4 — COPILOT() Function: AI-Powered Financial Commentary (Preview)

Generate Financial Commentary Directly in Cells

⏱️ 20 minutes 🎯 AI-in-cell for analysis & narrative 📄 Workbook: Finance-Copilot-Labs-4-6.xlsx

Why This Matters

Until now you've used Copilot through the side pane. The new COPILOT() function puts AI directly inside a cell — chainable with PIVOTBY, TEXTJOIN, FORMULATEXT, and everything else in Excel. For finance, this is transformative: imagine generating variance commentary, categorizing transactions, or rating risk — all as live formulas that update when data changes.

🔴
Preview Feature: COPILOT() is currently in Preview. You need a Microsoft 365 Copilot license and the right update channel (Beta / Current Channel Preview). If you get #NAME?, the function isn't available in your Excel build yet.

Open the Workbook

  1. Open Finance-Copilot-Labs-4-6.xlsx from your OneDrive.
  2. Navigate to the "VarianceReport" sheet tab.
  3. Review the P&L data — Actual, Budget, Variance $, Variance %, and Prior Year columns.

4.1 Generate a P&L Variance Summary

Start with a high-level narrative summary of the entire P&L. Click into an empty cell below or to the right of the data.

📋 Formula — type in an empty cell =COPILOT("Analyze this profit and loss statement and write a 3-sentence executive summary of the most notable variances. Focus on what a CFO needs to know.",A1:H30)

What to expect

  • A narrative paragraph in a single cell summarizing the key variance story
  • References specific line items with actual dollar and percentage figures
  • Written in a professional finance tone ready for a management package

4.2 Generate Line-by-Line Variance Commentary

Every month, finance teams write commentary explaining why each major line item is over or under budget. This is the single biggest time-sink in the close process. Let Copilot draft it.

📋 Formula — type in an empty cell =COPILOT("For each line item, write a one-sentence variance explanation suitable for a monthly management report. Focus on items where the variance exceeds 5%. Format as: [Line Item]: [explanation]",A1:H30)
💡
Key insight: The commentary updates automatically when the underlying data changes. Next month, paste in new actuals and the commentary regenerates — no rewriting needed.

4.3 Structured Multi-Section Analysis

For the monthly finance package, you need organized commentary by section. Combine COPILOT() with TEXTJOIN() for clean formatting.

📋 Formula — type in an empty cell =TEXTJOIN(CHAR(10)&CHAR(10),TRUE,COPILOT("Analyze this P&L and write a brief commentary organized into these sections: 1) Revenue Performance, 2) Gross Margin Analysis, 3) Operating Expense Highlights, 4) Bottom Line Assessment. Each section should be one short paragraph.",A1:H30))

What to expect

  • A structured, multi-section analysis in a single cell
  • Each section separated by a blank line (thanks to TEXTJOIN + CHAR(10))
  • Ready to paste into a memo, email, or presentation

4.4 Rate Variance Risk Level

Now let's get Copilot to assess which variances pose the most risk — useful for triage in close meetings.

📋 Formula — type in an empty cell =COPILOT("Rate each line item's variance risk as Low, Medium, or High based on the magnitude of variance % and dollar impact. Return the line item name and risk rating.",A1:H30)

4.5 Ask Copilot to Explain Its Own Formula

This is a powerful learning technique — ask Copilot to explain what a formula does.

Go to the "VarianceReport" sheet. Click on any cell that has a Variance % formula. Note its cell reference (e.g., F4). Then in an empty cell:

📋 Formula — type in an empty cell =COPILOT("Explain in plain English what this formula is doing and why it might show a #DIV/0! error",FORMULATEXT(F4))
💡
Teaching moment: This is incredibly useful for onboarding new team members or auditing inherited workbooks. Instead of deciphering complex formulas manually, ask Copilot to explain them.

Instructor Talking Point

The COPILOT() function represents a fundamental shift: AI is no longer a sidebar tool — it's part of the spreadsheet fabric. For finance teams, this means variance commentary, risk ratings, and even draft audit notes can be generated as formulas that stay live and update with the data. Emphasize: this is the future of the monthly close.

⚠️
Important: Never use Copilot output as the sole basis for a financial decision without independent verification. AI-assisted analysis is a productivity tool — not a substitute for professional judgment. Always review and validate before including in official reports.

Self-Check — Lab 4

Lab 5 — Month-End Close: Bank Reconciliation & Exceptions

Reconcile, Match, and Flag Exceptions

⏱️ 15 minutes 🎯 Reconciliation, matching & close tasks 📄 Sheets: BankRec, CloseChecklist

Why This Matters

Bank reconciliation is the gatekeeper of the monthly close. Until the bank rec is clean, you can't close the books. It's also one of the most manual, error-prone tasks in accounting — matching hundreds of transactions between two sources, hunting for timing differences, and chasing down discrepancies. Copilot can accelerate the matching and surface exceptions that need human investigation.

Time comparison

⏳ Manual: 60–90 minutes per account for matching and investigation ⚡ With Copilot: 15–20 minutes for initial matching and exception flagging

Open the Sheets

  1. In Finance-Copilot-Labs-4-6.xlsx, switch to the "BankRec" sheet tab.
  2. You'll see two data sections: Bank Statement (left) and General Ledger (right). These represent the two sides of the reconciliation.

5.1 Identify Unmatched Transactions

The first step in any bank rec is finding what doesn't match between the two sides.

📋 Prompt Compare the Bank Statement transactions with the General Ledger transactions. Identify items that appear on the bank statement but not in the general ledger, and items in the general ledger that don't appear on the bank statement. Match by amount and approximate date.

What to expect

  • A list of unmatched items from each side
  • Bank-side-only items are typically: bank fees, interest, items not yet recorded
  • Book-side-only items are typically: outstanding checks, deposits in transit

5.2 Flag Amount Discrepancies

Sometimes transactions exist on both sides but with slightly different amounts (data entry errors, transpositions).

📋 Prompt Find transactions where the bank statement and general ledger have similar descriptions or dates but the amounts don't match exactly. Show the bank amount, book amount, and the difference.
💡
Tip: These discrepancies are often caused by transposition errors (e.g., $1,530 vs. $1,350). In a real close, these are exactly the items your staff accountant would need to investigate and post correcting entries for.

5.3 Check for Duplicate Entries

Duplicates in either the bank statement or the general ledger create reconciliation nightmares.

📋 Prompt Check the General Ledger data for any duplicate entries — transactions with the same date, amount, and similar description. Highlight any duplicates found.

5.4 Summarize the Reconciliation Status

📋 Prompt Summarize the bank reconciliation status: total bank statement balance, total general ledger balance, identified reconciling items, and the remaining unreconciled difference.

What to expect

  • A reconciliation summary with both balances and reconciling items
  • Clear identification of outstanding checks and deposits in transit
  • Any remaining difference that needs investigation

5.5 Review the Close Checklist

Switch to the "CloseChecklist" sheet. Let's use Copilot to analyze close readiness.

📋 Prompt Analyze this month-end close checklist. How many tasks are complete, in progress, not started, and blocked? Which blocked tasks pose the highest risk to the close timeline? Summarize the overall readiness status.
💡
Real-world application: Controllers run this analysis every day during close week. Having Copilot summarize status and flag risks turns a 15-minute review into a 30-second prompt.

Instructor Talking Point

This lab demonstrates Copilot's ability to compare datasets — a fundamental accounting skill. While Copilot won't replace your reconciliation tool (Blackline, Trintech, etc.), it can serve as a quick first-pass analysis or work on ad hoc reconciliations where those tools aren't configured. The key message: Copilot handles the comparison mechanics; you bring the accounting judgment.

Self-Check — Lab 5

Troubleshooting & Common Issues

Issue Solution
Copilot button doesn't appear in Excel Ensure the file is saved to OneDrive or SharePoint (not stored locally). Copilot requires cloud storage. Also verify your Copilot license is active and the file is in a supported format (.xlsx).
Copilot says "I can't work with this data" The data must be in a formatted Excel Table (not just raw cells). Select the data range and press Ctrl+T to convert it to a table first.
"Insert Column" or "Apply" buttons don't appear Wait for Copilot to fully process the request. If the buttons are missing, try rephrasing the prompt to be more explicit about the action (e.g., "Add a column called X" instead of "Calculate X").
COPILOT() function returns #NAME? The COPILOT() function is in Preview. You need to be on the Beta or Current Channel Preview update channel. Check File → Account → Update Channel. Labs 0–3 and 5 work without it.
Variance calculations seem wrong (revenue shows as unfavorable when over plan) Variance direction depends on the formula. For revenue: positive variance is favorable. For expenses: negative variance (under budget) is favorable. Verify by checking the formula Copilot generated.
Aging buckets are incorrect for paid invoices Rephrase your prompt to explicitly handle paid invoices: "If Amount Paid equals Invoice Amount, show 'Paid' — otherwise, calculate the aging bucket based on Days Outstanding."
Copilot generated commentary cites wrong numbers Verify the cell reference range in your COPILOT() formula covers all the data. If the range is too narrow, Copilot only sees a subset. Expand the range and re-enter the formula.
Bank rec matching is imprecise Copilot matches on heuristics (amount, date proximity, description similarity). For production reconciliations, use dedicated recon tools. Copilot is best for quick ad hoc analysis and first-pass matching.

Discussion Questions

Use these for group debriefs in instructor-led sessions, or reflect on them individually:

  1. Which lab exercise maps most directly to a task you do every month? How much time would Copilot realistically save you on that specific task?
  2. In Lab 4, the COPILOT() function generated variance commentary. What review steps would you add before including AI-generated commentary in an official finance package?
  3. Think about your month-end close process. Which manual steps could Copilot handle, and which require human judgment that shouldn't be delegated to AI?
  4. How would you train a new staff accountant to use Copilot effectively? What guidelines or guardrails would you set for your team?
  5. What data do you work with regularly that isn't in a table format? How could you restructure it to work better with Copilot?

Key Takeaways

01

Real Formulas, Not Magic

Every column Copilot adds uses real Excel formulas you can inspect and edit. It's a productivity accelerator, not a black box.

02

Prompt Quality = Output Quality

Specific prompts with column names, thresholds, and output formats get dramatically better results than vague requests.

03

COPILOT() Changes the Game

AI as a native Excel formula means commentary, categorization, and risk ratings that stay live and update with your data.

04

Trust but Verify

Copilot handles the mechanics. You bring the accounting judgment. Always review AI output before it enters official records.

05

Compound Time Savings

One prompt saves minutes. A full workflow (GL prep → variance → aging → commentary) saves hours every close cycle.

06

Start with Your Workflow

The best way to learn Copilot is to apply it to your actual data. Start with the task you dread most each month.