There is a number that says more than any product launch presentation. Fewer than 4.5% of Microsoft 365's 450 million users decided to pay for Copilot. Not because they did not know it existed, but because it did not convince them enough to open their wallets.
Microsoft has known this for some time. This week, they said it out loud.
An internal memo from EVP Jacob Andreou, leaked in early July, revealed that the Copilot team, roughly 11,000 people, received an unusually blunt directive: the product must earn the right to exist. This is not the usual language from the company that dominated the productivity market for three decades. It is an acknowledgment that being installed on every screen is not the same as being chosen.
The result is a restructuring arriving this August: one single Copilot for everything, agents that work in the background without you asking, and the removal of features that collected digital dust.
The Number No Product Launch Wants to Show
The 4.5% is not just low. It is striking, given the context: Copilot has been embedded in Word, Outlook, Teams, and Excel for nearly three years. It comes pre-installed on every enterprise Windows machine. Nobody had to search for it or download it.
And yet, out of every 100 people with access, 95 kept doing things the old way.
The gap becomes clearer when compared to GitHub Copilot, the developer-focused version of the same product. That product converted: 4.7 million paid subscribers on a much smaller base. The difference is that developers could measure exactly how much time they saved. A programmer knows if they finished faster. An office worker rarely gets that same moment of certainty.
That is the gap Microsoft needs to close, and it is the real problem behind everything that is coming.
The One That Acts Without Being Asked
The boldest bet in the redesign is called AutoPilot. It is not an improved chatbot. It is a different category: AI agents that operate autonomously in the background, monitoring your email, calendar, and files, and taking action when they detect something needs attention.
The first to arrive is called Microsoft Scout. It can coordinate meetings across multiple attendees, summarize email threads, and manage repetitive workflows without the user initiating each task manually. It runs within the permissions the user already has inside Microsoft 365.
AutoPilot will carry an additional cost on top of the base Copilot price. The exact price has not been announced. The technical reason is straightforward: a chatbot query requires one model call, while an agent task may require between 10 and 20. The computational cost is substantially higher, and that gets passed on.
Everything in One Place, or How Microsoft Wants You to Stop Being Confused
One of the most consistent user complaints is that nobody knows what is what. Is Copilot in Outlook the same as the one in Teams? Does GitHub Copilot have anything to do with Copilot in Word? What is the difference between Copilot Chat and Copilot Cowork?
The honest answer is that they are different products sharing a name and not much else. Microsoft acknowledged this and decided to merge them.
The internal project goes by the name Copilot Fusion. The goal is a single app with one identity, where the same login works for GitHub and Microsoft 365, and where users switch between personal and work contexts within the same window. GitHub Copilot, Copilot Chat, Cowork, and AutoPilot will all live under the same roof.
The launch is targeted for August. Configuration details for enterprise environments have not been officially published yet.
What Is Being Cut, and Why It Matters
Two features were the first to go: Copilot Podcasts and Copilot Labs.
Copilot Podcasts auto-generated audio summaries of meetings and documents. Copilot Labs was the company's experimental features channel, where ideas were tested before being integrated into the main product. Both were discontinued for low user engagement.
What matters is not what disappeared. It is what this signals. It is the first time Microsoft has publicly acknowledged that Copilot features can be cut if they do not work. The implicit promise that everything that makes it into the product stays has expired.
Andreou's memo reformulated the standard: each feature must justify itself against the question of whether it helps complete real, measurable work. If it does not, it is out.
The Point
This August is not an interface update. It is a declaration that the original strategy, embed Copilot everywhere and wait for people to use it, did not work. Microsoft's answer is to consolidate, cut what does not function, and bet on autonomous agents as the mechanism that will finally create that "this actually saves me time" moment that users never had with the chatbot.
If you use Microsoft 365 at work, in the coming weeks you will see a different Copilot than the one you had. Understanding what changed before it arrives is worth your time.
Want to keep reading about AI at work, real productivity, and everything happening in the Microsoft ecosystem? Subscribe to the macareno.net newsletter. No spam, no filler.
Sources
- Microsoft Copilot Merges Into One App in August - TechTimes
- August 2026 Partner Center announcements - Microsoft Learn
- What's New in Microsoft 365 Copilot: July 2026 - A Guide to Cloud & AI