Microsoft 365 Copilot review: is it worth it?

Here's the short version of this Microsoft 365 Copilot review: it's worth it for a Microsoft-native team that will actually use it, and an expensive idle line item for one that won't. The product is genuinely good. The reason so many companies feel let down by it isn't the software; it's that the seats get bought and then barely touched. Copilot can't save time it's never asked to. So the real question isn't whether Copilot is capable, it's whether your team will be trained and supported to use it, which is the variable that decides the return. Copilot sits among the best AI tools for work when it's used well, and underperforms its price when it isn't.

With that framing set, here's what it does, where it shines, where it falls short, what it really costs, and who should buy it.

Microsoft 365 Copilot is an AI layer woven into Word, Excel, PowerPoint, Outlook, and Teams, plus a standalone Copilot app and chat. What sets it apart from a general chatbot is grounding: with a paid license it draws on your own tenant data, your email, files, chats, and meetings, while respecting existing permissions. In 2026 it added Agent Mode, which lets Copilot work iteratively inside a document rather than answering in one shot, along with named agents like Researcher and Analyst and an Excel agent that reached general availability that spring. A free, web-grounded Copilot Chat tier comes with eligible subscriptions; the paid license is what enables the in-app, data-grounded, agent-building capabilities that make it more than a novelty.

Copilot is an add-on, not a standalone product, and that matters for the math. The enterprise add-on runs about $30 per user per month, but it stacks on top of a base Microsoft 365 license, so the true all-in cost lands somewhere around $66 to $87 per user per month at the enterprise tier, and it scales with headcount rather than with outcomes. The small-business tier sits lower, recently around $21 per user with a promotional rate that's worth checking, since these figures move. Whatever the exact number on the day you buy, the most expensive version of Copilot is the one nobody uses, because then you're paying the full per-seat price for zero return. If you want the head-to-head against Google's option, our Gemini vs Copilot comparison covers it; this review stays on Copilot itself.

The sharpest criticism of Copilot is that you're blaming the customer for a product problem: the UK government ran a real trial, handed people the tool, and found no measurable productivity gain, so isn't that the product? Look closer at that same trial and the answer appears. Users averaged barely more than one Copilot action per working day while reporting 72% satisfaction, which is the exact signature of seats that were issued and then left idle, liked in principle and unused in practice. Analyst estimates put average activation around 8% of paid seats by the third month, which pushes the effective cost per active user well above list price. Training doesn't manufacture ROI out of a weak tool; it recovers the ROI a paid-but-idle tool is already costing you. That makes Copilot's disappointment an adoption problem, not a software one, and it's why getting a team to actually learn Copilot on their real work is what flips the verdict.

Buy Microsoft 365 Copilot if your work already lives in Microsoft 365 and you'll fund the adoption to go with it: train each role on real tasks, and the in-flow drafting, summarizing, and recap genuinely save time. Hold off if the seats will sit idle, or if you're a Google-native shop where bolting Copilot onto the wrong suite means fighting your own stack. The verdict, restated: Copilot is worth it conditional on use. The spend is decided the day you buy it; the return is decided by whether anyone is trained to use it. Get a team that actually uses it and the math works; skip the training and you've bought an expensive icon.