Microsoft hits 98% Copilot monthly usage

Microsoft’s Copilot story inside its own sales machine is interesting for one reason — it cuts against the usual AI rollout fantasy. Big company buys licenses, turns on assistant, productivity magically appears. But six months after Microsoft launched Copilot across its 62,000-person Microsoft Customer and Partner Solutions group, that is not what happened. Usage lagged. Then, over roughly two years, Microsoft changed the operating model and got to more than 60% daily usage and more than 98% monthly usage. ### Why is this a real story? Because MCAPS is not a pilot. It is one of Microsoft’s biggest sales organizations, and the case was published on May 12, 2026 through Harvard Business Review’s Cold Call podcast around a teaching case on Microsoft’s deployment of Copilot and agents. This is Microsoft basically saying: even we did not get broad transformation from launch-day excitement alone. (hbr.org) ### What broke in the first rollout? The gap was not access. Employees had the tool. The gap was practical use. People did not automatically know where Copilot fit into actual sales work, so the early excitement did not convert into durable habits. Microsoft’s case description says the company had to evolve beyond the standard deployment playbook and lean into focused training, peer champions, and habit building. (hbr.org) ### What changed after that? Microsoft stopped treating Copilot like a generic assistant and started wiring adoption into roles, workflows, and distribution. That matters more than it sounds. In enterprise software, discovery is half the battle. If workers have to guess which AI tool is safe, useful, and approved, most will just not bother. Microsoft’s own Copilot Studio team made the same point this week — the big blockers are where to start, what to trust, and how to get people to use agents in daily work. (hbr.org) ### Why does a curated store matter? Because abundance is the problem. Once a company has dozens of agents, “more choice” starts to feel like an unmoderated app store inside the firewall. Microsoft’s answer is the Agent Store — a curated hub in Microsoft 365 Copilot where users can discover agents from Microsoft, partners, and internal teams, with IT controls over what shows up. Basically, it turns agent adoption from a scavenger hunt into a governed catalog. (techcommunity.microsoft.com) ### Where do Copilot Studio and the SDK fit? They are the plumbing. Microsoft says organizations can populate that store in four main ways: deploy prebuilt agents, build agents in Copilot Studio, bring existing agents in, or integrate more deeply with the SDK stack. The point is not the menu itself. The point is that enterprise adoption needs a distribution layer and a governance layer at the same time. Without both, every new agent becomes another compliance review and another change-management headache. (techcommunity.microsoft.com) ### What is the “Sales Agent” angle? This is the next step up the ladder. Copilot helps in the flow of work. Sales Agent, as described in the HBR case, can take action on its own within preset guardrails and handle end-to-end customer interactions. That is a much bigger organizational leap, because now the question is not just “will employees use AI?” but “what work should AI do autonomously, and under whose rules?” (techcommunity.microsoft.com) ### So what does 98% monthly usage actually mean? It does not mean every seller loves Copilot every hour of every day. Monthly active usage is a broad measure. But in a 62,000-person sales organization, 98% is still a huge signal. It says the tool moved from optional novelty to normal infrastructure. And the more revealing figure may be the 60%+ daily usage — that suggests habit, not just occasional experimentation. (hbr.org) ### Why should other companies care? Because this is the part of enterprise AI people keep underestimating. The hard part is not model access. It is operational design — training, trust, curation, guardrails, and role-specific workflows. Microsoft’s own internal story makes that pretty clear. If one of the world’s most AI-native software companies needed two years and a lot of organizational scaffolding to make Copilot stick, everyone else probably does too. (hbr.org) The bottom line is that Microsoft’s 98% figure is less a victory lap than a reality check. AI adoption at scale looks a lot less like a product launch and a lot more like running an internal platform. (hbr.org)