
Microsoft announced multiple enterprise AI and cloud updates at the Ignite 2026 conference in San Francisco on November 17, U.S. time, repositioning Copilot as a work operating system and simultaneously launching the Microsoft IQ enterprise contextual platform and Agent 365 cost analysis tool. This round of updates reflects that the focus of enterprise AI applications has shifted from simple question-answering to cross-system task execution, requiring IT departments and management to address issues such as autonomous agent permission isolation, cross-platform knowledge integration, and out-of-control computational expenses.
Copilot integrates into Office and collaboration spaces; most agent functionalities initially limited to Frontier trial
At the productivity application level, Microsoft announced the integration of Copilot Home for conversation and task handover, the Office in Copilot that directly embeds Word, Excel, and PowerPoint into the interface, and the Copilot Managed Runtime, responsible for hosting employee-created applications, officially launched as of today. The official emphasized that documents or spreadsheets created within Copilot are included in existing subscription licenses without additional usage charges.
As for the more anticipated multi-party and autonomous agent features, they are still in the early access phase. The Copilot Workspaces, evolved from Copilot Notebooks, allows team members to share a working environment with AI agents, and the Autopilot, which continues to operate in the background and actively follows up on Teams channels and emails, is currently available only to customers who have joined the Frontier early access program; the Today feature, which integrates calendars and to-do lists, is in private preview, while the Copilot Code, which develops dashboards and tracking tools using natural language, will be officially launched in the coming weeks. Regarding third-party system connections, the Salesforce in Copilot plugin has entered public preview and includes over 38 sales skills, supporting business processes alongside more than 100 default skills in Dynamics 365.
In the announcement, Microsoft cited several enterprise adoption statistics, stating that nearly 80% of Fortune 500 companies worldwide have used Cowork, the Volkswagen Group is equipping over 330,000 employees with Copilot, and The Estée Lauder Companies is also extending its application to over 14,000 employees globally. However, Microsoft did not disclose specific cost savings or return on investment data for these companies after implementation.
Microsoft IQ unifies enterprise knowledge; cross-platform support for third-party AI assistants
To address the issue of AI agents repetitively creating business contexts across different departments, Microsoft has officially launched the enterprise contextual management platform Microsoft IQ, integrating Work IQ, Fabric IQ, Foundry IQ, Web IQ, and over 350 connectors, with no separate charging items. Along with business plugins in the Frontier program, department experts only need to organize business rules and permission settings once for the organization’s agents to reuse, while also supporting GitHub Copilot, Microsoft Foundry, and external assistants like Claude and ChatGPT through the Model Context Protocol (MCP) and API, reducing the risk of being locked to a single assistant.
Microsoft stated that nearly 60,000 customers are currently using Microsoft IQ, with the advertising group Publicis Groupe combining the platform, industry knowledge, and proprietary data to build custom AI skills and conduct real customer scenario testing. On the underlying model side, the small model Microsoft-Decision-1.1, which supports multimodal functionality and is responsible for workflow routing and classification, has been officially launched in Foundry, with more than 4,000 customers adopting it; the agent program development models Code 2 and Code 2 Flash were officially launched in GitHub Copilot, and Foundry has also introduced GPT-6.2 and Meta's latest models.
Agent 365 incorporates FinOps accounting; Perception and Windows enhance agent security
As autonomous agents gain direct access to enterprise systems, spending and security governance have become core concerns for CIOs and CFOs. Microsoft pointed out that the Agent 365 control plane now manages over 50,000,000 agents, and the newly established Agent 365 Business Analytics (launched in the Frontier public phase) attempts to link AI spending with task success rates and saved working hours, assisting management in verifying actual output value; the Agent 365 Cost Management, still in the preview phase, allows IT teams to track usage by user, service, and model, setting spending limits and alerts for Cowork to avoid budget overruns.
On the security protection level, Microsoft Perception, previously known as Project Perception, has been officially launched, collaborating according to scripts with agents driven by models such as MAI-Cyber in red teams (simulated attacks), blue teams (threat detection), and green teams (vulnerability remediation), retaining a manual approval mechanism for high-impact operations. In endpoint and identity management, Windows provides more refined operating system layer isolation through Microsoft Execution Containers (MXC), with each Autopilot agent assigned an independent Microsoft Entra Agent ID and designated manager; all logins and operations are audited, allowing IT departments to block abnormal agents at any time. Microsoft cited that financial institution BNY is exploring Perception capabilities in conjunction with Microsoft Defender, Entra, and Purview, and has previously collaborated with Microsoft on the first U.S. Treasury bond transaction participated in by an AI agent; at the same event, Microsoft also initiated an AI transparency initiative with over 40 organizations and NIST.
Azure Cobalt 200's performance increased by 50%; first round of 14 regions does not cover Hong Kong
In response to AI agents frequently needing to run tools, compile code, and retrieve files beyond reasoning, Microsoft has officially launched the next-generation Arm architecture Azure Cobalt 200 virtual machines, utilizing TSMC's 3-nanometer process with a 132-core processor, paired with Azure Boost and Azure Integrated HSM hardware security modules. Official test data shows that the overall CPU performance of Cobalt 200 has improved by up to 50% compared to the previous Cobalt 100, with database workload and AI agent tool execution performance improving by up to 1.9 times and 1.5 times, respectively, and introducing high-memory Mpsv4 and high-capacity storage Lpsv5 series.
However, for IT teams of enterprises in Hong Kong, regional support is still awaiting expansion. Cobalt 200 is currently only officially available in 14 Azure regions during its initial phase, covering only Eastern Australia, Central Indonesia, and Central South Korea in the Asia-Pacific region, with the East Asia (Hong Kong) region not included in the first round; Microsoft expects to expand to 23 regions by the end of the year and plans for Cobalt 200 to become a major part of Azure's added computing capacity by 2027.
Cloud development and infrastructure tools have also been updated, including the Azure Catalyst, which integrates Azure and GitHub agent tools to assist with modernizing legacy applications and databases (with the Azure Catalyst agent officially launched), the managed PostgreSQL database Azure HorizonDB, and the enterprise data application platform Fabric apps have all been officially launched. Microsoft also announced its self-developed Maia 300 chip, claiming a 40% higher output per watt than mainstream GPU systems.
Source: Microsoft Source Asia, The Official Microsoft Blog, Microsoft Copilot Blog, Microsoft Azure Blog, Azure Cobalt 200 Blog
This article is reprinted from unwire.pro, with copyright held by the original author and original media.
免责声明:本文章仅代表作者个人观点,不代表本平台的立场和观点。本文章仅供信息分享,不构成对任何人的任何投资建议。用户与作者之间的任何争议,与本平台无关。如网页中刊载的文章或图片涉及侵权,请提供相关的权利证明和身份证明发送邮件到[email protected],本平台相关工作人员将会进行核查。