Share via

MS Copilot Chat vs Agents outputs

Rodrigo Maldonado 0 Reputation points
2025-11-19T23:13:04.57+00:00

Hi I'm trying to build an agent in MS Copilot Agents but I can't reproduce with the same quality the output I'd expect if I compare the MS Copilot Agent from MS Copilot Chat using GPT-5. I wonder if Agents is using an inferior model and if there's a chance to use GPT-5 instead. I'm trying that Agents could generate charts for an executive summary but it can't, while Chat does it very fluently. Any suggestions?

Microsoft Copilot | Microsoft 365 Copilot | Development

2 answers

Sort by: Most helpful
Most helpful Newest Oldest
  1. Sayali-MSFT 5,361 Reputation points Microsoft External Staff Moderator
    2025-11-21T09:46:50.75+00:00

    Hello Rodrigo Maldonado,
    Copilot Chat generally runs on GPT-4 Turbo or GPT-5, making it more capable for creative and analytical tasks, while Copilot Agents use a lighter model focused on workflow orchestration, reliability, and integration with enterprise systems. This difference explains why Chat produces richer summaries and visuals, whereas Agents deliver more structured and procedural outputs.

    Microsoft does not currently allow switching Agents to GPT-5, as model selection is controlled internally. To work around this, you can combine Chat with Agents, extend Agents through external APIs, or break tasks into smaller steps so an Agent can generate data while another service produces the charts or visuals.

    1 person found this answer helpful.
  2. Q&A Assist
    2025-11-19T23:13:22.5666667+00:00
    AI generated content. This question contains an answer created with AI from Q&A Assist. Learn more

    The difference in output quality between Microsoft Copilot Chat and Microsoft 365 Copilot Agents can be attributed to the underlying models and the way they are configured. Microsoft 365 Copilot Chat is powered by the latest models, including GPT-5, which allows for more fluent and sophisticated responses, especially in generating content like charts for executive summaries.

    On the other hand, the agents in Microsoft 365 Copilot may not utilize the same advanced models or configurations, which can result in less fluid outputs. If you're experiencing limitations with the agent's ability to generate charts, it may be due to the specific capabilities and instructions set for that agent.

    To improve the output quality of your agent, consider the following suggestions:

    1. Adjust Agent Instructions: Refine the instructions or prompts used in your agent to better align with the desired output. This can help the agent utilize its capabilities more effectively.
    2. Utilize Declarative Approaches: If possible, use the declarative approach to build your agent, which allows for more tailored instructions and knowledge integration.
    3. Test and Iterate: Continuously test the agent with various inputs and refine its instructions based on the outputs you receive. This iterative process can help enhance its performance over time.

    Currently, there is no direct option to switch the model used by the agents to GPT-5, as they are designed to operate within their own framework and capabilities. However, ongoing updates and improvements may enhance their functionality in the future.

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.