Hands-on: an end-to-end workflow across Excel and Word

What you'll build

You're going to produce a short sales summary report for Priya, going from raw data to finished document in one connected workflow:

In Excel: clean and analyse the Fernway sales data with Copilot.
The handover: verify the findings before they leave the spreadsheet.
In Word: turn the verified findings into a written summary with Copilot.
The final check: confirm the report matches the numbers.

Set aside about 20 minutes. The task assumes a Microsoft 365 Copilot licence for the in-app features, but there's a free-chat fallback at each step, so you can complete the whole thing without one. Nothing here uses confidential data; the Fernway files are safe to practise on. If you swap in your own material, keep the Phase 0 privacy rules and strip anything sensitive first.

Before you start

Download the Fernway sales data. It's deliberately messy (blank totals, region typos, a wrong £49 total that should be £490).
Open it in Excel and, if you have Copilot, make sure the Copilot button is on the ribbon. No licence? Keep the free Copilot chat open instead; you'll paste data and findings into it.
Have a blank Word document ready for the report.

Step 1: Clean and analyse in Excel

First, make the data a proper table (select it, Insert → Table) so Copilot has clean structure. Then audit it before trusting any total.

Audit before analysingCopilot

This table is our April to May sales. Before any analysis, find data problems: blank Units or Total cells, Region values that look like typos, and rows where Units × Unit Price doesn't equal Total. List each with its row.

Why this works: It points Copilot at specific, checkable faults (blanks, typos, totals that don't reconcile) so you fix the data before it drives any conclusion. Analysis on dirty data is confidently wrong, so this step comes first.

Fix what it finds (the region typos, the £49 total, the blanks you can recompute). Then analyse:

Analyse the cleaned dataCopilot

Now, using the corrected figures: give me total sales by Region and by Product, say which region and which product performed best, and note any obvious trend between April and May. Keep it to a short bulleted list of findings.

Why this works: It names the exact cuts (totals by region and product, plus the trend) so you get the specific findings the report needs, not a generic overview. A named ask is a checkable ask.

Free-chat fallback: paste the CSV rows into the free Copilot chat and run the same two prompts. You won't get a live PivotTable, but you'll get the audit and the findings as text, which is all you need for the report.

Step 2: The handover, verify before you carry it across

This is the step people skip, and it's the whole point of the lesson. A wrong number that leaves Excel becomes a wrong number in the report, then a wrong number in Priya's inbox. Before you copy anything into Word:

Spot-check two or three figures against the actual rows. Does "North strongest" hold up if you glance at the North totals?
Confirm the correction stuck: is the £490 in, and are the region typos merged so "Sotuh" isn't counted as its own region?
Write the findings down as a short, plain list you're confident in. That list, not the raw sheet, is what goes to Word.

Step 3: Turn findings into a report in Word

In Word, use Draft with Copilot and feed it your verified findings, not the raw spreadsheet.

Draft the report from verified findingsCopilot

Draft a one-page sales summary for Priya, Head of Operations, from these findings: [paste your verified list]. Structure: a two-sentence headline, then 'By region', 'By product', and 'Trend' sections, then one line on data quality noting we corrected some errors before analysis. Plain, professional tone, about 250 words. Don't add any numbers beyond the findings I've given you.

Why this works: It hands Copilot the checked findings and sets audience, length and structure, so Copilot writes the report rather than re-deriving numbers it might get wrong. Drafting from your verified list keeps the data trustworthy across the handover.

That last instruction, don't add any numbers beyond the findings I've given you, is a guardrail against Copilot inventing plausible detail to round out the report. Then refine as normal: "tighter, and lead with the headline figure", "add a sentence recommending we watch the South region".

Free-chat fallback: the free chat drafts from pasted findings exactly the same way. You lose the in-document editing, but the report comes out identical. Paste your findings and the same prompt.

Step 4: The final check

Before this is "done", read the report against your findings list one more time. Every number in the report should match a number you verified in Step 2. Copilot may have rounded, reworded a comparison, or slipped a figure; catch it here, not after you've sent it.

Your checklist

[ ] I formatted the sales data as a table before asking Copilot anything.
[ ] I ran the audit and fixed the region typos, the £49→£490 total, and the blanks.
[ ] I got findings by region, product and trend from the corrected data.
[ ] I spot-checked two or three figures against the rows before leaving Excel.
[ ] I drafted the Word report from my verified findings, not the raw sheet.
[ ] I refined the draft at least once.
[ ] I read the final report against my findings and every number matched.

Self-check

Did a bad number survive to the report? Trace the report's headline figure back through Word → your findings → the Excel rows. If it holds all the way, your handovers worked. If it broke somewhere, notice where; that's the weak link to watch next time.
Did Copilot invent anything in the write-up? Compare the report to your findings list. Any number or claim in the report that isn't in your list is something Copilot added, a hallucination to cut, however plausible it reads.
Could you repeat this without the instructions? The pattern of analyse, verify, hand over, draft, verify again is the reusable skill. If you could run it on a different dataset tomorrow, you've got it.

Take it further

Common mistakes

Skipping the audit and analysing dirty data. The wrong totals and region typos flow straight into your conclusions. Clean first, always.
Carrying the raw sheet into Word instead of verified findings. Draft from a checked list, not the spreadsheet, so Copilot writes about numbers you trust rather than re-reading ones you don't.
Letting the report add numbers. Without the "don't add anything beyond my findings" guardrail, Copilot can pad a report with invented specifics. Name the guardrail and check for breaches.
Treating the handover as a copy-paste, not a checkpoint. Each app boundary is where an error can slip through unnoticed. A ten-second spot-check at each one is the habit that makes the whole workflow safe.
Over-trusting the finished report because it reads well. A fluent, well-structured summary feels finished, but polish is not accuracy, and Copilot is at its most error-prone with the numbers this report is built on. The final trace-back against your verified findings is the only real check; never send it just because it looks right.

Keeping current

The apps and buttons will change, but the workflow won't: analyse, verify, hand over, draft, verify again. That chain-with-checks is the durable skill this whole level builds towards. For the current state of Copilot across the apps, Microsoft's Copilot help and learning hub is the place to look. Accurate as of 13 July 2026.

By the end, you'll be able to…

Why it matters