Microsoft sidelined its AI chief on March 17. Mustafa Suleyman—the DeepMind co-founder hired to lead Copilot—got reassigned to “superintelligence” research, a polite way of saying he’s no longer running the product. Jacob Andreou, a former Snap executive with no public AI wins on his résumé, now runs the entire Copilot organization. The reason? Copilot’s web market share sits at 1.1%, down from 1.5% a year ago, while Gemini grabbed 21.5% and Grok stole half of Microsoft’s remaining share in a single month.
This isn’t a typical reorg. It’s an admission that selling AI to enterprises and getting humans to actually use it are two completely different problems—and Microsoft just bet its AI future on fixing the latter by demoting the visionary who couldn’t ship a product people wanted to open twice.
Microsoft’s 1.1% web share problem has a $30-per-seat solution nobody’s buying
The adoption crisis is real. Microsoft claims 15 million paying Copilot users—up 160% year-over-year, impressive on paper. But only 3.3% of Microsoft 365 users who try Copilot Chat actually pay for it. That’s not a conversion problem. That’s a product-market fit problem.
The web traffic tells the real story. Copilot’s share dropped 17% quarter-over-quarter from October to December 2025, according to Seoprofy’s analysis. Enterprises are buying seats to check an AI compliance box, but employees open Copilot once during onboarding and never return. The gap between “deployed” and “used daily” is where Microsoft’s AI strategy is quietly dying.
Meanwhile, AI adoption in high-skill jobs continues to accelerate—just not with Copilot. Google’s Gemini is winning the consumer war Microsoft never seriously contested, and now the enterprise war looks shakier than the earnings calls suggest.
15 million paid seats sound impressive until you see the denominator
Microsoft has roughly 400 million Microsoft 365 commercial seats. That means Copilot penetration after 18 months of availability is about 3.75% of the installed base. The “90% of Fortune 500 use Copilot” talking point—if it’s even accurate—obscures what “use” means. Most are pilot programs or IT-mandated rollouts generating zero organic demand.
The math gets worse. Microsoft reports 33 million total active Copilot users across all platforms. If 15 million are paying enterprise customers, that leaves 18 million free or consumer users generating that dismal 1.1% web share. For context, ChatGPT and Gemini each have hundreds of millions of users. Microsoft isn’t competing in the same weight class.
And here’s the thing: reports suggest Microsoft’s own engineers don’t use Copilot for daily work. When your internal teams avoid the product you’re selling to the world, the adoption theater becomes harder to sustain.
Andreou inherits a product roadmap with no public wins—and Suleyman’s exit kills momentum
Jacob Andreou’s LinkedIn lists roles at Snap, but no named product launches that would explain why he’s now responsible for Microsoft’s most important product bet. That’s not necessarily disqualifying—plenty of great product leaders don’t have flashy portfolios. But it is risk.
The bigger problem? Suleyman wrote in an internal memo that he’d been “thinking about how we” should evolve the strategy—corporate speak for “this wasn’t working and everyone knew it.” His reassignment to superintelligence research means the person who understood the original vision is now irrelevant to quarterly execution. Enterprises betting on immediate Copilot ROI just inherited 6-12 months of alignment tax while the new team figures out what product they’re actually building.
The shift toward voice AI replacing typing at work requires agents that anticipate needs, not chatbots that wait for prompts. That’s the transition Microsoft is failing to execute, and leadership chaos doesn’t accelerate product velocity.
Microsoft can’t win the consumer AI war—Gemini and ChatGPT already own that. But it can’t afford to lose the enterprise war either, because that’s where the recurring revenue lives. Andreou’s job isn’t to make CIOs buy Copilot. They’re already buying it. His job is to make employees want to use it after IT deploys it. The question isn’t whether Microsoft can sell Copilot to enterprises. It’s whether it can sell Copilot to the people enterprises are buying it for.
