Microsoft Merged All Its Copilots Into One App — But Almost Nobody Is Paying

There is a version of the Microsoft Copilot story that is straightforwardly impressive: the company took a fragmented product line — consumer Copilot here, enterprise Copilot there, different apps with different capabilities — and unified it into a single experience with a wave of genuinely useful features. That version is true. There is another version, told by the same reporting, where the company that dominates the productivity software market cannot convince its own customers to pay for or regularly use the AI it built into that software. That version is also true. Both versions are the same story.

What did Microsoft actually ship?

By August 2026, Microsoft completed the merger of every consumer and enterprise Copilot into a single unified app. The product fragmentation that defined Copilot’s first two years — different Copilots for different surfaces, with different capabilities and confusing naming — is over. One app, one brand, one experience across consumer and business use.

The feature list that came with the merger is substantial:

1. Hey Copilot — hands-free voice activation

   A voice command that launches Copilot without touching a keyboard or clicking an icon. Modeled on the “Hey Siri” and “OK Google” patterns, but wired into the full Copilot capability set — not just search, but document creation, data lookups, and workflow automation.

2. Copilot Vision — on-screen and image analysis

   Copilot can now read and analyze content visible on screen and in images, providing contextual answers faster. Instead of describing what you need help with, you can point Copilot at what you are looking at and ask questions about it directly.

3. PowerPoint intelligence that respects your brand

   PowerPoint Copilot now reuses existing themes and branding when generating or modifying slides. It understands images and shapes in context, letting users ask it to explain or adjust visuals while keeping slide styling on-brand — a direct response to the early complaint that AI-generated slides looked generic.

4. Faster content ingestion via parallel crawls

   Copilot connectors now run content crawl and identity crawl in parallel rather than sequentially, making ingested organizational content available to users significantly faster. This matters for enterprise deployments where Copilot needs to understand a company’s internal documents.

Microsoft also formally announced Guild Copilots — a new tier of industry-specific AI assistants, following successful pilots in healthcare and legal. This signals a strategic bet that the next wave of enterprise AI adoption comes from specialization: an AI that knows medical terminology and clinical workflows, not just “an AI that can summarize documents.”

What is the uncomfortable number behind all of this?

Here it is, plainly: fewer than 4.5% of Microsoft’s 450 million commercial Microsoft 365 customers currently pay for Copilot features. And of those who do pay — that small slice of an enormous customer base — only 20–30% use it on a weekly basis.

Run the math and the picture sharpens. Microsoft has roughly 450 million commercial 365 seats. Fewer than 4.5% paying for Copilot means fewer than about 20 million paid Copilot users. Of those, 20–30% using it weekly means somewhere between 4 and 6 million people are actually making Copilot a regular part of their work. Out of 450 million.

Metric	Figure
Commercial Microsoft 365 customers	~450 million
Paying for Copilot features	Fewer than 4.5%
Weekly active use among paid users	20–30%
Implied weekly active Copilot users	~4–6 million
Single unified Copilot app	Launched August 2026
Guild Copilots announced	Healthcare and legal pilots

Why is the gap between “shipped” and “used” so wide?

The honest answer is that nobody fully knows yet, but the pattern is consistent enough to identify the forces at work.

The price signal is new. Microsoft 365 customers already pay for their productivity suite. Copilot is an additional cost on top of that — and justifying incremental spend requires demonstrating incremental value that shows up in something a CFO can measure. “It helped me draft an email faster” is hard to put in a spreadsheet. The value is real but diffuse, and diffuse value loses budget fights to things with clear ROI line items.

Habits are harder to change than features. People who have used Word, Excel, and PowerPoint the same way for a decade do not spontaneously start talking to an AI sidebar because a feature update lands. The feature has to interrupt an existing workflow at exactly the right moment — and the AI has to be right often enough, fast enough, that the interruption feels like help rather than friction. That is a behavior-change problem, not a technology problem, and behavior-change problems are slower.

The product was fragmented until now. Before the August 2026 unification, Copilot was a confusing landscape of different products with overlapping names. If you were a Microsoft 365 admin trying to evaluate “Copilot” for your organization, you had to figure out which Copilot, with which capabilities, for which users, at which price point. The merger fixes this going forward, but it means the first two years of Copilot’s life were spent building adoption on shifting ground.

What is Microsoft doing about it?

The August updates read, in hindsight, as a direct response to the adoption problem — not just new features for their own sake, but features designed to lower the barriers that kept the numbers low.

The parallel connector crawl improvement is easy to overlook, but it may matter more than the flashier features. One of the most common complaints about enterprise AI tools is that they do not “know” enough about the company to be useful — and the time between deploying Copilot and having it actually understand your organization’s documents was a real adoption barrier. Making that window shorter directly addresses the “I tried it and it didn’t know anything” first impression that kills habitual use.

Why this matters for anyone watching the enterprise AI market

This is not a story about Microsoft failing. It is a story about the gap between enterprise AI feature velocity — how fast companies can ship impressive capabilities — and enterprise AI habit formation — how fast humans actually start using those capabilities as part of their daily work.

That gap is the defining challenge for every company selling AI into the enterprise right now. The technology is, by any reasonable standard, impressive. Microsoft is shipping features that would have been science fiction five years ago. But the adoption data shows that shipping features and changing behavior are two different problems, and the second one is much harder and much slower.

The bottom line

Microsoft is doing two things at once: building one of the most capable AI assistants on the market and discovering, in public, that capability alone does not drive adoption. The August 2026 merger and feature wave are a direct response to that discovery — a company that realized its AI product had a usage problem and is systematically trying to fix it by removing friction, adding specificity, and making the first experience faster.

Whether it works is the most important open question in enterprise AI right now. Not because Microsoft’s future depends on it — 365 is dominant with or without Copilot — but because if the company with the largest installed base of productivity software, the deepest enterprise relationships, and functionally unlimited AI investment cannot get habitual usage past single-digit percentages, it tells us something fundamental about how fast AI actually changes the way people work. The answer, so far, is: slower than the feature releases.