Global ideas, clearly explained.

Technology

Microsoft Rebuilds Copilot Around Home, Code and Autopilot, Betting the Work Surface Is the Moat

Microsoft has rebuilt Copilot around three capabilities β€” Home, Code, and Autopilot β€” turning the AI assistant into a unified work surface where users chat, build apps in natural language, and delegate to persistent agents billed by usage.

MMeridian NewsroomAuthor
30 Sept 2026, 2:41 amUpdated 30 Sept 2026, 2:54 am5 min read0 views
Microsoft Rebuilds Copilot Around Home, Code and Autopilot, Betting the Work Surface Is the Moat
An editorial illustration of a futuristic unified workspace interface with three glowing zones β€” Home, Code, and Autopilot β€” representing Microsoft's Copilot rebuild.Credit: AI-generated editorial illustration β€” Meridian

On September 25, 2026, Microsoft unveiled a fundamental rebuild of Copilot, its AI productivity platform. The redesigned app is organized around three capabilities β€” Home, Code, and Autopilot β€” and it represents the company's most aggressive attempt yet to answer a question that has dogged its AI strategy: what, exactly, is Copilot for? The answer, increasingly, is everything you do at work.

The redesign moves Copilot beyond a chat sidebar attached to Office apps. Microsoft is positioning the new app as a unified work surface: a place to converse with AI, build internal tools by describing them in plain language, and delegate recurring work to agents that keep running when you're away. It is also a pricing experiment β€” splitting AI spending between fixed user subscriptions and usage-based billing for agentic work.

Home: chat, Cowork, and Office in one place

Home is the new default starting point, combining Copilot Chat with Cowork, Microsoft's collaborative delegation feature. The headline change is Office in Copilot: full versions of Word, Excel, and PowerPoint are built directly into the experience, so users can create and edit real documents, workbooks, and presentations alongside a conversation β€” without switching applications.

The consolidation addresses a long-standing complaint. Since Microsoft began layering Copilot across Windows, Teams, and Office separately, users and IT teams alike struggled with which Copilot product did what. Folding them into one app with one default screen is Microsoft's answer to that confusion.

Code: build apps by describing them

Code is the most striking of the three capabilities. Powered by the same underlying technology as GitHub Copilot, it lets non-developers build apps, trackers, dashboards, automations, and workflows by describing what they need in natural language β€” from desktop widgets to cloud-hosted internal apps shared with a team. Microsoft says Code runs in a sandboxed environment and can be hosted within an organization's own Microsoft 365 tenant, with Copilot Managed Runtime providing IT-governed hosting for these apps.

The implications for IT departments are significant. "This is Microsoft now bringing in coding agents, AI coding agents right into a knowledge worker's experience, which is kind of bizarre or will be an unusual thing for CIOs to see," Gartner analyst Larry Cannell told TechTarget. "The coding tab and giving access to an AI coding agent is an industry experiment." Sandbox controls, plugin approvals, and governance of what these generated apps can touch become first-order concerns the moment the rollout reaches real tenants.

Autopilot: the agent that keeps working

Autopilot β€” previously known as Scout, introduced at Microsoft's Build developer conference earlier this year β€” is the persistent, always-on digital coworker. Microsoft describes it as a proactive agent with its own identity, memory, computer, and workspace inside a customer's tenant, able to set up and manage review processes, build schedules, prepare meetings, and follow up, across Teams and Outlook, subject to permissions, auditing, and governance and usage-based pricing.

Notably, Autopilot agents run continuously in the cloud under their own governed identity β€” Entra ID β€” and users can choose which model powers the work: OpenAI's GPT models, Anthropic's Opus, an automatic router, or xAI's Grok, with IT admins able to restrict which model families specific groups may use. The multi-model stance is a quiet admission that no single lab's models will win every enterprise workload.

The billing shift: Copilot Credits

The operational change may matter more than the interface one. Autopilot, Code, and Cowork move to usage-based billing through Copilot Credits, separate from the standard per-seat Copilot license, which covers chat and Office integration. Microsoft is drawing a clearer line between two kinds of AI spending: predictable fixed subscriptions for everyday use, and metered consumption for long-running agentic work.

The split also adds FinOps controls β€” dashboards for tracking AI spend β€” acknowledging that autonomous agents running unattended in the cloud can generate surprising bills. For organizations evaluating Microsoft 365 Copilot, the redesign changes not just the product but how it is budgeted, deployed, and governed.

Why Microsoft is doing this now

The context is competitive pressure and uneven adoption. Reports suggest fewer than 20 million users pay for Copilot out of more than 450 million Microsoft 365 customers, and enterprise buyers have been unsure what the product was supposed to be. Meanwhile, rivals have spent the past year racing toward agentic workflows: Anthropic's Claude models, OpenAI's agents, and even SpaceX's acquisition of the AI coding company Cursor have all signaled that the industry sees coding agents and persistent assistants as the next battleground. Microsoft's own September was crowded β€” ChatGPT in Word, Gemini integrations, and new agent billing from Perplexity all landed within days of each other β€” which makes the consolidation play as much defensive as offensive.

Microsoft's response is characteristically platform-shaped: don't sell a better chatbot, sell the work surface where chat, code, and agents live together β€” and where the documents, meetings, and identity systems of the enterprise already are. The moat is not the model; it is the tenant.

Availability and open questions

The rollout is staged. Home and Code begin rolling out through Microsoft's Frontier early-access program in the coming weeks, with Autopilot entering private preview at the end of September 2026. Broader previews for Microsoft 365 Premium and Pro subscribers are planned later in 2026, though Microsoft has not announced a general availability date for full agent collaboration features.

Open questions remain: how usage-based pricing lands with CFOs, whether IT teams will embrace sandboxed code generation from non-developers, and whether Autopilot's promise of an agent with its own identity survives contact with enterprise security reviews. What is not in question is the direction β€” Microsoft is betting that the future of work software is not an assistant you ask, but a surface where agents work alongside you.

Sources

Frequently asked questions

What are Home, Code, and Autopilot in the new Copilot?

Home combines chat, Cowork, and Office in Copilot (full Word, Excel, PowerPoint in one place). Code lets users build apps, dashboards, and automations by describing them in natural language. Autopilot is a persistent always-on AI agent for delegated, recurring work.

When will the redesigned Copilot be available?

Home and Code begin rolling out through Microsoft's Frontier early-access program in the coming weeks; Autopilot enters private preview at the end of September 2026. Broader previews for Microsoft 365 Premium and Pro subscribers are planned later in 2026.

Can Autopilot use AI models other than Microsoft's?

Yes. Users can choose OpenAI's GPT, Anthropic's Opus, an automatic router, or xAI's Grok, with IT admins able to restrict which model families specific groups may use.

How is the new Copilot billed?

The standard license covers chat and Office integration. Autopilot, Code, and Cowork bill separately through usage-based Copilot Credits, alongside new FinOps dashboards for tracking AI spend.

M

Written by

Meridian Newsroom

The Meridian newsroom desk. AI-assisted drafts are always reviewed by a human editor before publication.

Comments

Comments are moderated. Please keep the discussion civil.

Loading comments…

Related stories