Microsoft Copilot After 6 Months: What Actually Works (And What's Theater)

Three enterprise rollouts over six months. About 800 seats total. Healthcare, financial services, professional services. Same product, same vendor, same Microsoft sales pitch. Three very different outcomes.

Here's what I'd tell anyone considering Copilot before they spend the budget.

What actually earns its keep

Copilot in Outlook for email triage and drafts. This is the killer feature. Long thread → tight summary. Vague reply request → competent draft. Calendar question → reasonable proposed time. Of all Copilot's surfaces, Outlook is where it pulls real weight every day.

Copilot in Teams meeting summaries. Not perfect, but solid. Action items get captured. Decisions get flagged. Late joiners can catch up in 90 seconds. The fact that it's integrated and the recording flow is one click makes it sticky.

Copilot for SharePoint search. When your team's knowledge is buried in 14 sites and 12,000 documents, Copilot can find what older keyword search couldn't. This is genuinely transformative for orgs with substantial internal content.

Excel Analyze Data + Copilot prompting. For data work the user can describe in plain English ("what's the trend in customer count by region year over year"), Copilot reliably produces the right pivot or chart. Faster than building it manually.

Word for first drafts of repetitive document types. Memos, status updates, project briefs from notes — Copilot drafts and you edit. Works.

What's theater

Copilot in PowerPoint generating slides. The output looks like 2019 stock-deck Microsoft. Layouts are flat. Images are awful. You'll redo most of it. Save your money and time.

Copilot summarizing short emails. "This email is about a meeting." Yes, thanks. The minimum-useful-context for summarization is a 5+ message thread. Below that it's pointless.

Copilot in OneNote for note-taking. Nominally works, in practice everyone is using Notion or Apple Notes anyway. Adding Copilot doesn't fix that.

Copilot "Pages" as a new productivity surface. Microsoft's bet on a new collaborative canvas. Most teams I've seen don't adopt it. They stay in Word + Teams.

Copilot in Edge as a browsing assistant. It exists. Almost nobody uses it. Chrome wins the browser war regardless of AI features.

The actual ROI math

At $30/user/month for Copilot Pro, you need each user to save about 1 hour/month at their loaded cost to break even. For knowledge workers at $80k+ loaded comp, that's a low bar.

In practice, the users who use Outlook + Teams summaries + SharePoint search heavily get 3-5 hours/month back. They love it. They're the easy ROI case.

The users who try to use it everywhere (PowerPoint, OneNote, Edge, custom workflows) burn time on the features that don't work and have a net-negative experience.

The split inside an organization is real. Some seats are 10x ROI. Some seats are net-negative. The average looks fine but hides the spread.

Where Copilot wins over ChatGPT Enterprise

Three places:

1. Data is already in Microsoft 365 (SharePoint, OneDrive, Outlook, Teams). Copilot operates on it natively without copy/paste.

2. Compliance posture for regulated industries. Microsoft has more enterprise-grade data handling for many use cases. Easier audit story.

3. IT integration. Your team already has Microsoft licensing, admin tools, governance. Adding Copilot is incremental.

Where ChatGPT Enterprise wins over Copilot

Three places:

1. Quality of long-form generation. ChatGPT (and Claude) consistently produce better drafts than Copilot for serious writing tasks.

2. Custom GPTs / project workflows. More flexible than Copilot's templated agents.

3. Pricing for high-volume use. ChatGPT Enterprise's pricing math can work out better for power users.

The deployment pattern that works

For most clients I've helped, the right Copilot rollout looks like:

Phase 1 (Weeks 1-4): Targeted pilot with 30-50 users who are heavy Outlook + Teams users. Get the easy wins on email and meeting summaries.

Phase 2 (Weeks 5-12): Expand to users with heavy SharePoint dependency. Train them on search and Excel use cases.

Phase 3 (Months 4-6): Broad rollout with explicit guidance on what to use vs what to skip. Most teams that skip this phase end up with frustrated users and quiet attrition.

The orgs that just turn it on for everyone without training and selection see disappointing ROI. Not because the product is bad — because the use case mismatch is real.

My one rule

If you're considering Copilot: deploy it. The integration with Microsoft 365 is real value.

If you're considering Copilot AND ChatGPT Enterprise: deploy both. They're not competitors for most users. They're complementary tools for different jobs.

If you're deciding between Copilot and "no AI yet": Copilot wins. Inside an MSFT enterprise, the integration alone justifies the $30 even at a fraction of the use cases working.

What I'd skip

Don't pay for Copilot for users who don't live in Outlook + Teams + SharePoint daily. Field staff who mostly use mobile or specialized apps won't get value. Buy seats based on actual usage, not titles.

Don't bother with the productivity-suite-wide rollout pitch from your Microsoft account team. Pilot, measure, expand. Don't trust the licensing math without your own usage data.

Don't expect Copilot to replace your specialized AI tools. Sales teams still need Gong/Outreach. Engineering teams still need Cursor/Claude Code. Copilot is general-productivity AI, not specialist AI.

The bottom line

Copilot earns its $30 when used correctly. Microsoft hasn't done a great job teaching users which use cases work and which don't. You have to figure that out yourself — or have someone who's already done it tell you.