Microsoft 365 Copilot has moved from a headline feature to something millions of office workers now touch every day, whether they realize it or not. At Topeny, we spent three weeks running Copilot through a battery of everyday office scenarios inside Word, Excel, Outlook, and Teams to see whether it genuinely saves time or simply adds another layer of friction to already busy workflows. This test report walks through our methodology, what we found in each application, where Copilot excelled, where it stumbled, and whether it is worth the added subscription cost for a typical small-to-midsize business.
Test Methodology
Our testing team used a licensed Microsoft 365 E3 tenant with the Copilot add-on enabled across a group of ten test accounts representing different roles: two executives, three analysts, two customer-facing sales reps, two administrative staff, and one IT operations lead. Each role was given a set of realistic tasks pulled from real workplace requests we have collected over the past year, ranging from drafting a client proposal to reconciling a messy expense spreadsheet. We logged time-to-completion, accuracy of the output, number of manual corrections required, and overall satisfaction on a five-point scale after each task.
We ran every task twice: once with Copilot assistance and once without, using the same person for both runs on different days to reduce the learning-curve effect. This gave us a reasonably clean before-and-after comparison rather than a single subjective impression.
Word: Drafting, Rewriting, and Summarizing
Copilot in Word performed best when given a clear structural request, such as “turn these five bullet points into a two-paragraph executive summary” or “rewrite this section in a more formal tone.” In these cases, our testers reported a roughly 40% reduction in first-draft time compared to writing unassisted. The tone-matching feature, which lets you specify formal, casual, or persuasive language, was consistently reliable across dozens of trials.
Where Copilot struggled was in long-document consistency. When asked to summarize a 40-page contract, it occasionally missed clauses buried in later sections or slightly paraphrased legal terms in ways that changed their meaning. We do not recommend relying on Copilot for legal or compliance-critical summarization without a full human review pass. For internal memos, meeting notes, and marketing copy, however, it performed admirably.
Notable Word Findings
- Draft generation from bullet points: strong and consistent
- Tone and style rewriting: highly reliable
- Long-document summarization (30+ pages): inconsistent, missed details in three of ten trials
- Grammar and clarity suggestions: on par with or better than the built-in Editor tool
Excel: Formulas, Data Cleanup, and Chart Generation
This is where opinions among our testers diverged the most. Copilot in Excel can generate formulas from plain-language prompts, such as “calculate the month-over-month growth rate for column C,” and it got the formula syntax right in the large majority of test cases. It also proved genuinely useful for identifying anomalies in messy datasets, flagging duplicate rows and inconsistent date formats that a human might scan past.
However, when we asked it to build a pivot table summarizing regional sales trends from a 12,000-row dataset, Copilot needed two rounds of clarification before producing the correct grouping, and one of our analysts noted it initially miscounted a subtotal by including a header row in the sum. This is a meaningful caveat: financial and reporting teams should treat Copilot’s Excel output as a fast first draft, not a final number to present to stakeholders without verification.
Chart generation from natural language (“show me a bar chart comparing Q1 and Q2 revenue by region”) worked smoothly and saved real time, especially for staff who are not naturally comfortable with Excel’s charting menus.
Outlook: Email Drafting and Inbox Triage
Copilot’s “Summarize this thread” feature in Outlook was one of the most consistently praised features across our entire test group. For long email chains with ten or more replies, it condensed the conversation into three or four accurate sentences in nearly every trial, saving several minutes per thread for busy managers.
Draft reply suggestions were helpful for routine correspondence (scheduling, confirming details, polite decline emails) but felt generic for anything requiring nuance, such as addressing a client complaint. Testers still had to significantly rewrite Copilot’s suggestions in sensitive situations, which somewhat offset the time savings for that category of email.
Performance and Reliability
Response latency averaged between 3 and 9 seconds for most requests, which felt acceptable in daily use but noticeably slower than typing manually for very short tasks. We experienced two outright failures during our three-week window where Copilot returned an error instead of a response, both during periods of apparent high demand on Microsoft’s infrastructure. Neither failure resulted in lost work, but it is worth setting expectations that Copilot is not 100% available on demand.
Pricing
As of our testing window, Microsoft 365 Copilot is sold as an add-on to qualifying Microsoft 365 business and enterprise plans, priced per user per month, with an annual commitment typically required for the best rate. For a 20-person team, this represents a meaningful line-item increase to the existing software budget, so the return on investment really depends on how document- and email-heavy a given role is. Roles that spend hours per week drafting, summarizing, or analyzing spreadsheets saw the clearest payback; roles that are mostly in meetings or doing hands-on physical work saw very little benefit.
Pros and Cons
| Pros | Cons |
|---|---|
| Strong drafting and rewriting quality in Word | Long-document summarization can miss details |
| Excellent email thread summarization in Outlook | Excel formulas need verification on complex datasets |
| Formula and chart generation saves real time in Excel | Per-user pricing is a real budget consideration |
| Deep integration across the whole Microsoft 365 suite | Occasional latency and availability hiccups |
| Tone and style matching is consistently reliable | Generic replies for emotionally sensitive emails |
Mobile App Experience
We also tested Copilot on the Word, Excel, and Outlook mobile apps for iOS and Android. The mobile experience is noticeably more limited than desktop: Copilot in Word mobile can rewrite and summarize selected text but does not currently support the full drafting flow available on desktop. On Outlook mobile, “Summarize this thread” worked reliably and was arguably the single most useful mobile AI feature we tested, since it let testers triage long email chains from a phone during commutes without needing to open a laptop. Excel mobile Copilot support was the thinnest of the three, largely limited to basic formula help rather than full pivot table or chart generation. Teams relying heavily on phones and tablets for work should treat Copilot as a desktop-first tool for now.
Security and Data Privacy Considerations
A common question from IT administrators we spoke with during testing was whether Copilot uses company data to train Microsoft’s underlying models. Based on Microsoft’s published documentation at the time of our testing, Copilot for Microsoft 365 commercial customers operates within the organization’s existing Microsoft 365 tenant boundary and respects existing permission and compliance boundaries such as sensitivity labels, meaning Copilot should not surface content a given user does not already have permission to see. IT administrators in our test group were able to enable or restrict Copilot access at the user and group level through the Microsoft 365 admin center, which several of them found reassuring for a phased rollout. Organizations in regulated industries should still review Microsoft’s current compliance documentation directly, since AI governance policies continue to evolve.
Onboarding and Training Time
One underrated factor in our results was how quickly each role became comfortable with Copilot. Testers who received a short 30-minute walkthrough of prompt-writing basics (being specific about tone, length, and format) produced noticeably better first-attempt results than testers who simply opened Copilot and started typing casual requests. We strongly recommend a brief, structured onboarding session rather than assuming staff will discover best practices on their own, since the gap in output quality between a trained and untrained user was one of the largest variables in our entire test.
Frequently Asked Questions
Does Copilot work with older versions of Office?
Copilot requires a qualifying Microsoft 365 subscription with cloud-connected apps; it does not work with older perpetual-license versions of Office such as Office 2019 or Office 2021 purchased as a one-time license.
Can Copilot see data outside my organization’s tenant?
In our testing, Copilot only drew on content the signed-in user already had permission to access within their own Microsoft 365 tenant, plus general web grounding when that feature was enabled by the administrator.
Is Copilot worth it for a small team?
Based on our results, small teams with document- and email-heavy roles saw a clear time benefit, while teams with mostly hands-on or meeting-heavy roles saw less return relative to the added per-user cost.
How accurate are Copilot’s Excel formulas?
Formula syntax was correct in the large majority of our trials, but we recommend manually verifying any formula used for financial reporting or decisions with real business impact.
Final Verdict
Microsoft 365 Copilot earns a solid 4 out of 5 from our testing team. It is at its best as a drafting and summarization assistant, and it genuinely reduces time spent on repetitive writing and email triage. It is at its weakest when asked to handle high-stakes accuracy tasks such as legal summarization or complex financial reporting, where human verification remains essential. For organizations already committed to the Microsoft 365 ecosystem and with document-heavy workflows, the add-on cost is likely to pay for itself within a few months. For lighter-usage teams, it may be worth piloting with a small group before rolling it out company-wide.
As always, Topeny recommends running your own pilot with a representative sample of your team’s actual daily tasks before making a purchasing decision, since results can vary meaningfully depending on document complexity and role.
Leave a Reply