The Front Door Has Moved

First, the web browser was supposed to kill the desktop suite. Then it was Google Docs. Then Slack was going to make email and document collaboration obsolete. Now the prevailing hot take is that autonomous agents and conversational AI will turn Word, Excel, and PowerPoint into relics of an analog corporate past.

If you only scan the headlines, it might sound like I’m about to call the time of death on Microsoft 365, too.

I’m not.

What we’re witnessing isn’t the death of the productivity suite. It’s a metamorphosis.

The applications aren’t disappearing. But their role in the daily workflow is undergoing its biggest inversion in decades.

The front door to productivity has moved.

From Model Routing to Work Routing

A few days ago, I wrote about Audiophile AI. My argument was simple: for 95% of business users, obsessing over which foundational model answers your prompt is like debating vinyl versus lossless FLAC through a pair of cheap laptop speakers.

Increasingly, we shouldn’t need to care.

Microsoft has already been moving in this direction at the model layer. Copilot Chat can automatically route prompts to different models depending on the task, using faster models for straightforward work and specialized reasoning models for more complex requests.

Microsoft’s introduction of Microsoft-Decision-1 reinforces the larger architectural pattern. It is a specialized decision model designed for things like routing, classification, prioritization, and workflow control. The idea is that software should increasingly be capable of deciding what should happen next.

Now, apply that same logic one layer higher.

For decades, productivity software has required humans to act as manual work routers.

Have an idea? First decide whether it belongs in Outlook, OneNote, Word, Loop, Excel, or PowerPoint. Launch the application. Find the information scattered across emails, meetings, documents, chats, and business systems. Move pieces between windows. Finally, start producing something.

We’ve become so accustomed to that process that we rarely stop to question it. But in an agentic workplace, starting with the application container increasingly feels backward.

If I shouldn’t have to choose the underlying model, why should I always have to choose the application before I’ve even articulated the business problem?

That’s the more interesting shift underway.

We’re moving from model routing to work routing.

How We Got Here

This transition didn’t happen overnight. When Microsoft introduced Microsoft 365 Copilot in March 2023, much of the attention naturally went to Copilot inside Word, Excel, PowerPoint, Outlook, and Teams.

But there was another piece of the announcement that was arguably more important to where we are today: Business Chat.

From the beginning, Microsoft described an experience that could work across Microsoft 365 applications and organizational data. Instead of starting inside a specific application, you could start with a question.

That was an early glimpse of the front door moving.

Later that year, Microsoft made Copilot generally available for enterprise customers and positioned Microsoft 365 Chat as an experience spanning emails, meetings, chats, documents, Microsoft 365 applications, and the web.

Then came Copilot Pages in 2024, providing a persistent, editable canvas for AI-assisted collaboration rather than leaving everything trapped inside an ephemeral chat.

Agents pushed things further. Instead of Copilot simply responding to prompts, Microsoft began expanding the platform around specialized agents capable of supporting business processes across applications and data sources.

Each step reduced the assumption that work must begin inside a traditional application. And in September 2026, Microsoft made that direction much more explicit.

The company described the new Copilot Home as a starting point for work, bringing Chat and Cowork together while also introducing Office in Copilot. Word, Excel, and PowerPoint become accessible from within that broader Copilot experience. Microsoft also previewed an experience where, eventually, you won’t necessarily choose between Chat, Cowork, or Code. Instead, you state what you want to accomplish and Copilot routes the work to the appropriate capability.

Think about the progression. Copilot started as intelligence embedded inside the applications. Increasingly, the applications sit behind Copilot as specialized environments that can be brought into the workflow when they’re needed.

Intent is becoming the interface.

The Bigger Question

That does not mean Word, Excel, or PowerPoint are going away. In fact, I think that framing misses the more interesting story entirely.

The question isn’t whether Copilot replaces Microsoft 365.

The question is what happens to Microsoft 365 when the application is no longer necessarily the place where work begins.

What happens when we stop starting with:

“Which application should I open?”

and increasingly start with:

“What am I trying to accomplish?”

That subtle change has enormous implications for how we create documents, analyze information, collaborate with teams, govern organizational knowledge, and design the digital workplace. And that’s where this gets really interesting. Because as the applications become less visible at the front of the workflow, the Microsoft 365 platform behind Copilot may actually become more important, not less.

That’s where I’ll pick up next.