When Microsoft first launched its Copilot for Microsoft 365, it wasn’t just introducing a new feature, it was setting the tone for enterprise AI adoption. With its deep integration into Word, Excel, PowerPoint, and other productivity staples, Copilot promised to revolutionize how knowledge workers operate. Fast-forward to today, and Microsoft has confirmed 8 million active paying users of Microsoft 365 Copilot. On paper, that’s a milestone. In context, it’s a warning.
According to reporting from Ed Zitron’s Where’s Your Ed At newsletter, those 8 million users represent just 1.81% of Microsoft’s 440 million Microsoft 365 subscribers. Even if we generously assume another 4 million paid licenses are sitting unused, a common scenario in enterprise software, that’s still only a 2.72% conversion rate. For a company with one of the largest commercial software empires in history, thousands of salespeople, and a global partner network, that’s a surprisingly low uptake.
Microsoft 365 Copilot isn’t cheap. At $30 per user per month (on top of an existing Microsoft 365 subscription), it’s a premium add-on. Zitron notes that Microsoft has already begun offering discounts, 15% off for bulk orders, promotional pricing through its Cloud Solution Provider program, and other incentives to drive adoption. Even with these efforts, the revenue from Copilot is estimated to be around $2.88 billion annually, a fraction of the $33.1 billion Microsoft earns each quarter from its Productivity and Business Processes segment.
So why isn’t Copilot catching on faster? Zitron cites a telling quote from The Information:
“It’s easy for an employee to say, ‘Yes, this will help me,’ but hard to quantify how. And if they can’t quantify how it’ll help them … it’s not going to be a long discussion over whether the software is worth paying for.”
That sentiment reflects a broader issue with generative AI in the workplace: the promise is abstract, the ROI is murky, and the actual usage is minimal. Microsoft defines an “active” user as someone who takes one action in any Microsoft 365 app using Copilot within 28 days. That’s a low bar for engagement, and it suggests that many users aren’t finding sustained value.
Zitron also reports that GPU utilization for enterprise Copilot is hovering around 60%, and that SharePoint, one of Microsoft’s most widely used business apps, had fewer than 300,000 weekly active users of its AI features in August. These numbers hint at a deeper problem: even when AI tools are available, they’re not being used consistently or meaningfully.
Microsoft’s Copilot rollout was supposed to be the enterprise AI success story. Instead, it’s become a case study in the limits of generative AI adoption. Despite aggressive marketing, deep integration, and a massive user base, the conversion rate remains low, the engagement shallow, and the revenue modest.
Zitron’s analysis is blunt: “Even Microsoft can’t sell AI.” And if the most powerful productivity software company in the world is struggling to make its AI assistant stick, what does that say about the rest of the industry?
Microsoft may have started the ball rolling on productivity AI, but the momentum hasn’t matched the hype. The Copilot story is still unfolding, but for now, it’s a reminder that adoption isn’t guaranteed, and that even the biggest players can stumble when the value proposition isn’t clear.
