Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
Your agents can use a deep reasoning model, the Azure OpenAI o1 model, to better understand and complete complex tasks. This capability enhances an agent's decision making and returns more accurate responses.
To use a deep reasoning model, add reason to a task or step in an agent's instructions.
You can add deep reasoning models to agents with generative mode and deep reasoning turned on.
Note
Adding a deep reasoning model in Copilot Studio is currently only supported in the US and EU (excluding UK) regions.
Important
This article contains Microsoft Copilot Studio preview documentation and is subject to change.
Preview features aren't meant for production use and may have restricted functionality. These features are available before an official release so that you can get early access and provide feedback.
If you're building a production-ready agent, see Microsoft Copilot Studio Overview.
How does deep reasoning work?
Deep reasoning uses the Azure OpenAI o1 model to complete tasks requiring logical reasoning, problem solving, and step-by-step analysis. This model allows your agent to complete complex tasks with greater accuracy, though with slower response times, than default models used my Copilot Studio.
Agents don't use deep reasoning automatically. Instead, you tell your agents to use deep reasoning to complete tasks by using the keyword reason in the agent instructions.
Turn on access to deep reasoning models
In order to use reasoning with an agent, you must turn on generative mode and access to deep reasoning models for that agent:
Open your agent.
Select Settings.
Select the Generative AI tab.
If your agent isn't already in generative mode, select Generative mode.
Turn on Use deep reasoning models.
Use a deep reasoning model to complete tasks
Agents don't use deep reasoning models automatically, or for all tasks or steps the agent performs. Instead, agents can use deep reasoning models to complete specific tasks.
To use deep reasoning, tell the agent to use reason in the agent instructions for completing a specific task or step.
For example, these instructions include the keyword reason in one of the steps. The agent uses the Azure OpenAI o1 model for that step only:
Parse the request body to identify: Location (e.g., Arizona) and services requested (e.g. cleaning, landscaping, maintenance).
Use Get suppliers from ERP flow to retrieve internal supplier data.
Search your internal KB for information on suppliers capacity, location, and performance.
Run search the web flow and extract Market Share, Revenue, Rank and other relevant information from all the suppliers.
Using the data, use reason to provide one top recommendation to the user request along with the relevant context.
Note
An agent can use deep reasoning models for multiple tasks or steps, but each one must contain the keyword reason. Because deep reasoning models take longer to respond, using models for multiple steps can substantially slow down your agent's response times.
Reviewing reasoning model behavior
You can trace when and how your agent uses deep reasoning models in the Activity page.
The activity map shows a deep reasoning node where the agent used a deep reasoning model.
To see details, select the reasoning node to expand it. The expanded node explains the reasoning steps and data the model used, and the result the model outputted.
