Smart feedback: continuous improvement for your conversational agents with the latest Microsoft features

Why feedback is essential in conversational AI

Conversational artificial intelligence solutions only generate real value if they evolve from the data they receive. That is why Microsoft has launched —and is launching— new capabilities in Copilot Studio and Dynamics 365 Customer Service that allow organizations to collect, analyze, and apply feedback in real time, with different approaches depending on the moment of the interaction.

In this article, we analyze three complementary mechanisms to enhance the continuous improvement cycle of conversational agents: two of them are in public preview and one is already generally available (GA). Knowing them and combining them properly can make the difference in the quality of the experience you offer.

1. Response feedback through Adaptive Cards (Copilot Studio)

One of the most relevant new features is the possibility of adding a feedback card at the end of each response generated by the conversational agent.

Copilot Studio has incorporated a new generative orchestration engine that allows actions to be triggered before reasoning a response, after reasoning it, or after sending the response to the user.

By leveraging these “triggers,” it is possible to send an Adaptive Card that collects feedback in a much more dynamic and interactive way, directly from the user after the response, to know if it was useful.

💡 This functionality is highly aligned with the principles of continuous learning in generative AI. The earlier improvement points are identified, the faster the agent can be trained to deliver more precise, clear, and personalized responses.

In addition, the use of Adaptive Cards facilitates visual and functional integration with multiple channels.

2. Funcionalidad de dedos arriba y abajo (Copilot Studio – GA desde el 9 de junio)

Recently, Copilot Studio has introduced a feature that allows you to enable a small visual feedback element on each response. It consists of two icons — “thumbs up” 👍 and “thumbs down” 👎 — which let users indicate whether their experience was positive or negative, along with a space to add comments.

The most interesting part is that these ratings can be stored in Dataverse, where an administrator can access them and use the data to improve development.

3. Post-service satisfaction surveys (Dynamics 365 – preview)

The Contact Center and Customer Service modules in Dynamics 365 make it possible to create and configure automatic post-interaction surveys that are sent at the end of a conversation with the agent (human or virtual).

These surveys are built with Copilot Studio and integrated with Customer Insights – Journeys, which allows:
• Personalizing the content and delivery channel.
• Measuring key KPIs such as CSAT or NPS.
• Triggering automated actions based on the survey result.

The results of these post-service surveys are stored together with the rest of Dynamics 365 data in Dataverse. This way, results can be viewed from the Contact Center or Customer Service interface, as well as analyzed with other solutions such as Power BI.

4. Developer feedback in Agent Builder (Copilot Studio)

Microsoft has recently incorporated a new mechanism inside Copilot Studio for developers to evaluate agent behavior during testing and leave comments directly on each response.

🔍 This feature is especially designed for development and testing teams, as it enriches the validation phase before deployment. It enables capturing qualitative perceptions that complement quantitative usage metrics.

A valuable tool to promote a culture of continuous improvement from the design stage and not only in production.

A continuous improvement strategy based on multiple sources

Each of these features responds to a different moment in the conversational journey.

Combined, they make it possible to build a 360° vision of conversational performance, identify friction points, and continuously improve quality.

What does it mean for two features to be in preview?

Incorporating tools in public preview implies accepting that they may undergo changes or improvements before general availability. However, it also represents a clear competitive advantage for companies seeking to stay ahead of the market, experiment with advanced capabilities, and prepare their conversational agents to deliver better experiences from the start.

At Bravent, we help you implement them safely and aligned with your business objectives.

Conclusion

Feedback in conversational AI is no longer optional: it is the engine that ensures the quality, usefulness, and efficiency of interactions. The new capabilities in Microsoft Copilot Studio and Dynamics 365 open the door to smarter, data-driven management with multiple observation points.

Adopting these tools is an opportunity to create conversational experiences that are more human, efficient, and aligned with the real expectations of users.

Do you want to incorporate these feedback capabilities into your conversational AI strategy with Microsoft Copilot Studio?

👉 Contact us at mailto:info@bravent.net and find out how to design more effective conversational agents from the very first message.