Custom engine agents for Microsoft 365 overview

Custom engine agents are conversational Teams bots designed to harness the powerful capabilities of Large Language Models (LLMs) for seamless user interaction. These advanced bots mark a significant departure from traditional bots, offering an extensive range of features that elevate the overall user experience.

A diagram that shows how a custom engine agent in Teams works

Custom engine agents utilize LLM technology that allows them to easily understand and respond to user queries, creating dynamic and immersive interactions. Custom engine agents also offer advanced functionalities such as UI manipulation, task execution, and content creation. These capabilities make custom engine agents indispensable tools for streamlining workflows and boosting productivity.

Important

  • Custom engine agents are conversational Teams bots with custom LLM, therefore they don't require Microsoft 365 Copilot license to build, debug and deploy.
  • The cost of a custom engine agent depends on the Azure services consumed in the app. To estimate the cost, visit the Azure Pricing Calculator.

Build custom engine agents

There are two options for building custom engine agents. You can use the Teams AI library and Teams Toolkit to build your agent with code, or you can use Microsoft Copilot Studio to build an agent with little to no code.

Teams AI library and Teams Toolkit

For enterprises seeking bot solutions that are precisely customized to user demands and scenarios, using the Teams AI library is a great choice. The Teams AI library utilizes LLMs that adhere to Microsoft's UX principles and that are compatible with Teams and Microsoft 365. Developing custom engine agents with the Teams AI Library can provide more natural, intuitive, and specialized user interaction experiences. This option is advantageous for developers wishing to customize model selection and orchestration, applying prior experience in Teams bot development, or aiming to ensure accessibility for all Microsoft 365 users.

Developers can also utilize the Teams Toolkit to construct their projects. The Teams Toolkit includes prebuilt templates for a seamless project launch, offering an ideal solution for quick starts, easy debugging, and deployment.

Explore templates to get started

Kickstart building your custom engine agents using Teams AI library with the prebuilt templates available in Teams Toolkit:

Template Description
Basic AI ChatBot Build a basic AI chatbot for Teams using Teams AI library.
AI Agent ChatBot Build an AI agent chatbot for Teams that can make decisions and perform actions based on LLM reasoning. AI Agent template has options to start from scratch or use OpenAI Assistants API.
Chat with your data Expand AI bot's knowledge with your content to get more accurate answers to your questions. Chat with your data supports Azure AI Search, Microsoft Graph Search with Microsoft 365 and SharePoint content, and custom OpenAPI and data ingestion options.

Take a tour of the sample scenarios

Explore sample scenarios available in Teams AI library GitHub repository to understand more about the capabilities you can add in your custom engine agents:

Sample Description C# JavaScript Python
List bot Harness the power of AI, simplify your workflow and bring order to your daily tasks with action chaining capabilities. C# JavaScript Python
DevOps bot Perform DevOps actions such as create, update, triage, and summarize work items. C# JavaScript Python
Twenty questions bot Use LLM capabilities and the concept of user intent by having users guess a secret within 20 questions, grounding the chatbot with some question and answer as a data source. C# JavaScript Python
Math tutor assistant Use OpenAI's Assistants API with the Code Interpreter tool and create an assistant that's an expert on math. C# JavaScript Python
Food ordering assistant Call actions to order food from a fictional restaurant called The Pub with complex interactions. C# JavaScript Python

Microsoft Copilot Studio

Microsoft Copilot Studio empowers teams to quickly and easily create powerful copilots using a guided, no-code graphical experience without needing data scientists or developers.

Note

Microsoft Copilot Studio refers to custom engine agents as custom copilots.