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.
The Fallback system topic triggers when a Copilot Studio agent doesn't understand a user utterance and doesn't have enough confidence to trigger any of the existing topics.
You can use the Fallback topic in many ways.
Offload single question and answer pairs
If you have a large number of question and answer pairs for single-turn conversations, consider offloading them to an external system and accessing them through the Fallback topic.
Copilot Studio limits an agent to 1,000 topics. You can decide to manage exceptions to the Copilot Studio topics that should generally do things for the user through the Fallback topic. By using the Fallback topic, your agent can get answers from an external system by calling a cloud flow from an action node.
Tip
- Question answering from Azure AI Language provides cloud-based natural language processing (NLP) that you can use to create a natural conversational layer over your data. It's used to find the most appropriate answer for any input from your custom knowledge base of information.
- Learn how to integrate question answering in Copilot Studio.
Give your agent a personality
Question answering from Azure AI Language provides a readily available "chitchat" data source, of various tones such as professional, friendly, witty.
Chitchat lets your agent answer trivial questions, such as "how are you feeling today, agent?"
Integrate with generative AI models
You can also use the Fallback topic to call a generative AI model, such as the models used by the Azure OpenAI Service, to handle more questions from your customers.
Through the system prompt and prompt engineering, you can also feed the model with instructions, rules, and input data before it generates an answer.
Learn from the Fallback topic
Track what your customers say that doesn't trigger a standard Copilot Studio topic. Use these phrases to enrich an existing topic's trigger phrases, especially when the user utterance should trigger a topic. These phrases can also lead to the creation of new topics to better serve your users as you better understand what they're trying to accomplish.