# Microsoft Adds Code Generation and Autonomous Agent to Copilot > Source: > Published: 2026-09-25 13:17:10+00:00 **September 25, 2026, (Inside AI)** — Microsoft on Friday announced a major expansion of its Copilot assistant, adding a natural-language coding tool and an autonomous enterprise agent. The move pushes the company deeper into territory held by specialized AI startups and signals a strategic shift from chat companion to full workplace platform. The new coding feature, called **Code**, lets users build applications, dashboards, and other software using plain English prompts. It runs on the same underlying technology as **GitHub Copilot**, Microsoft's developer-focused AI tool. Early-access customers will receive Code by the end of the month. Subscribers to **Copilot 365 Premium** and **Pro** tiers can preview it later this year. Microsoft also revealed **Autopilot**, a revamped version of the **Scout** agent it introduced in June. Autopilot enters private preview at the end of September. Unlike a standard chatbot, this digital coworker carries its own identity within a company's directory and operates under specific permissions that administrators control. The Redmond, Washington-based company is simultaneously embedding **Word**, **Excel**, and **PowerPoint** directly inside Copilot. Users can now collaborate on documents, spreadsheets, and presentations without switching between applications. The integration aims to reduce friction for office workers who currently toggle between multiple windows and tabs. ## Enterprise Trust Remains The Biggest Hurdle Security and governance concerns have slowed [agent adoption across large organizations](https://insideai.news/news/agentic-ai/ai-agent-trust/10072/). Microsoft's decision to give Autopilot its own directory identity and granular permissions addresses a specific pain point that has kept many enterprises on the sidelines. **"I think it's been very hard for many organizations to bring agents into the enterprise because of security, compliance and governance concerns," Annie Pearl, corporate vice president of Copilot Product, said in an interview. "So we've worked... to make sure that we were building this in a way that would really be able to be used in the enterprise."** Pearl's comments reflect a broader industry reality. AI agents that act autonomously on behalf of employees introduce risks that traditional software never posed. An agent with access to email, calendars, and internal databases could potentially leak sensitive information or take unauthorized actions. Microsoft's approach of assigning agents their own identities mirrors how companies manage human contractors and third-party vendors. The company also introduced cost-management tools that give employees direct visibility into expenses tied to their AI usage. This feature arrives as enterprises grow increasingly anxious about unpredictable AI bills. Unlike traditional software licenses with fixed monthly fees, AI assistants charge based on usage. Heavy adoption can lead to surprise invoices that strain IT budgets. Microsoft's push comes amid intensifying competition in the AI assistant market. **OpenAI's ChatGPT** and **Anthropic's Claude** have gained significant traction among knowledge workers. Both offer coding capabilities and agent-like features. Microsoft, which has invested billions in OpenAI, now finds itself competing directly with its own partner. The stakes extend beyond product features. Copilot represents Microsoft's primary vehicle for monetizing AI investments across its productivity suite. The company needs Copilot adoption to justify the massive capital expenditures it has committed to data centers and AI infrastructure. Each new capability serves dual purposes: attracting new subscribers and retaining existing ones who might otherwise defect to standalone AI tools. Code's rollout timeline suggests Microsoft is taking a measured approach. Early access begins this month, but general availability for premium subscribers stretches into late 2026. This staggered release allows the company to gather feedback and address security vulnerabilities before wide deployment. GitHub Copilot, which launched in 2021, faced early criticism over code quality and potential copyright issues. Microsoft appears determined to avoid similar problems with Code. Autopilot's private preview similarly reflects caution. By limiting initial access, Microsoft can test the agent's behavior in controlled environments. The company has not disclosed which organizations will participate or what specific tasks Autopilot will handle. Details about pricing for these new capabilities also remain unclear. The embedded Office applications represent a quieter but potentially more impactful change. By keeping users inside Copilot, Microsoft reduces the cognitive load of context-switching. Workers can ask Copilot to draft a document, then edit it directly without opening Word separately. This seamless experience could become a significant differentiator against competitors that lack native productivity integrations. Industry analysts have long predicted that AI assistants would eventually [absorb traditional software interfaces](https://insideai.news/news/agentic-ai/ai-agents-consumer-shift/12654/). Microsoft's announcement brings that prediction closer to reality. The question now is whether enterprises will trust autonomous agents enough to deploy them broadly, and whether workers will embrace a single AI-powered hub for their daily tasks.