Microsoft's Efforts Around Agentive AI

The next wave of AI is upon us in the form of agentive AI. Whereas the first wave, generative AI, was about creating content, agentive AI is about automation of repetitive tasks to free up people to do other things.

AI agents are comprehensive software applications that can receive input, perceive their environment, reason in response to actions and data, and act autonomously to achieve those goals. They leverage artificial intelligence techniques, including machine learning, natural language processing and reasoning algorithms, to perform a wide range of tasks with minimal human intervention.  

Microsoft's most public effort surrounding agentive AI is Microsoft 365 Copilot Chat, released in January. It is an AI-powered chat assistant integrated into all of the Office 365. With Copilot Chat, Microsoft has evolved Copilot from a generative AI assistant to a more autonomous agentic AI system. It offers agents for sale in its Agent Store, which you access through Copilot.

Navigate to the Copilot interface within your application, such as Teams and Microsoft 365's Copilot app, etc.) and select Get Agents from the right-side panel to access the Agent Store. The new "Agent Store" offers both Microsoft and partner-developed agents, as well as custom agents created by organizations.

The Agent Store for Office 365 has agents for:

• A new generation of agents built into Word, Excel, Outlook, Teams, and so on. They assist with summarizing documents, drafting emails and documents, analyzing documents and automating tasks like generating emails.
• Security and Compliance agents, a part of Microsoft Defender and Microsoft Purview services. They cover antimalware, attachment security, regulatory compliance, data loss prevention and information protection agents, and audit logs.
• Exchange Online agents, with mail flow rules, inspect and manipulate emails in transit, connection and content filtering, folder repair, running background processes on user mailboxes.
• Microsoft Entra (Azure AD) Conditional Access agents, which evaluate user sign-ins and enforce conditional access policies. They are used to restrict or allow access based on device compliance, location, or risk.
• Third-party monitoring and backup agents from vendors like Veeam and Barracuda that back up or monitor Office 365 data.

You can also build custom agents using no-code and low-code tools in Copilot Studio, an evolution of Power Virtual Agents. It's designed for users without programming skills to create, deploy and manage AI-powered agents.

Copilot Studio includes a visual, no-code interface and a natural language prompt-based authoring interface with more than 1,000 connectors to services. With the natural language prompting, users can describe what they want the bot to do in plain English. The system uses natural language understanding and generative AI to suggest dialog flows.

Dialogs (conversations) are built using a visual flowchart-like interface, where each step or "node" can represent a question, message, action, or condition. You can branch conversations based on user responses or logic.

Copilot Studio also comes with native support for Dataverse, SharePoint, Microsoft 365, and Azure services for things like data queries. It also has prebuilt, custom connectors to external data and services, such as Salesforce and ServiceNow.

In addition, Microsoft has a comprehensive collection of Agentive AI tools and technology on Azure perform certain functions or enable the development and deployment of intelligent agents to perform certain tasks.

Here's a summary of key Microsoft Azure services and initiatives relevant to agentive AI:

Azure AI Agent Service: This is a fully managed service specifically designed for building, deploying, and scaling AI agents. It handles the management of the infrastructure so developers can focus on agent logic and workflows. Automatic tool calling simplifies the process of enabling agents to interact with external systems and data sources.

Agent tools support integrating with services like Bing, Azure AI Search, and Azure Functions, as well as supporting custom tools via OpenAPI specifications. It offers enterprise-grade security with features like keyless authentication.

Azure OpenAI Service: This service provides access to language models like GPT-4o, GPT-4 and GPT-3.5 Turbo, offering coding and language AI models that you can fine-tune to your specific needs for a variety of use cases, not just agentive AI. These models can be used for natural language understanding, generation, and reasoning, which is needed for agents that interact with users or process information.

Azure AI Foundry: This initiative serves as a unified platform with APIs for models, tools, and workflows Customize, host, run, and manage AI-driven applications built in GitHub, Visual Studio and Copilot Studio. The Foundry has an extension for Visual Studio that allows developers to build, test, and deploy agent-based applications within Microsoft's IDE.

Azure Cognitive Services: This collection of AI APIs offers various capabilities that can be incorporated into agentive AI systems. They include natural language features like entity recognition, sentiment analysis, and key phrase extraction, which can help agents interpret user input and process textual data, speech-to-text and text-to-speech functionalities, image and video analysis capabilities, search capabilities that agents can use to access and retrieve information from various data sources and detection harmful or inappropriate content in user inputs and agent outputs.

Azure Digital Twins: Digital twins are simulations of real world entities, such as buildings, factories or even entire cities. The twins are continuously updated with real-time data for ongoing, evolving simulation over time. While this is not directly an agentive AI service, digital twins can provide a rich, contextualized environment for agents to operate within.

Microsoft's agentive AI offerings on Azure are part of a broader initiative to create autonomous and semi-autonomous AI agents that can perceive, reason, and act on behalf of users or systems. These offerings combine multiple Azure services, including Azure OpenAI, Azure Cognitive Services, Azure AI Studio, Azure Machine Learning and Power Platform, to build intelligent agents that perform tasks with minimal human intervention.

Posted by Andy Patrizio on 05/08/2025