Share via

AI-based agent authoring overview

Copilot Studio offers generative AI features to reduce manual authoring and dramatically expand the scope of an agent's knowledge and its ability to interact with users.

Generative AI is an artificial intelligence technology that uses language models to generate original content and provide natural language understanding and responses. Learn more about Generative AI in the Artificial Intelligence (AI) playbook.

In Copilot Studio, you can use the following generative AI features to retrieve and create content, either individually or all together.

  • Create an agent. With no manual authoring of topics required, an empty agent can generate answers based on knowledge sources you specify such as websites and files. Learn more in Quickstart: Create and deploy an agent.

  • Harness AI general knowledge. When Use general knowledge is turned on, the agent can answer general questions unrelated to your specific knowledge sources or topics. Learn more in Allow the agent to use general knowledge.

  • Author topics using natural language. Describe what you want your topic to do, and Copilot Studio creates it for you. Your agent includes conversational responses and multiple types of nodes. Use the suggested default topic as a starting point for further development. Learn more in Create and edit topics with Copilot.

  • Author prompts using natural language. Describe the prompt you want to create, and Copilot Studio generates it for you. You can use the suggested default prompt as a starting point for further development. Learn more in Create a prompt with Copilot.

  • Create agent flows using natural language. Describe the overall flow you want your agent to follow, and Copilot Studio generates it for you. You can use the suggested default flow as a starting point for further development. Learn more in Build an agent flow with natural language.

  • Use generative orchestration. Let the agent select the most appropriate topics, tools, agents, and knowledge sources at runtime. Learn more in Orchestrate agent behavior with generative AI.

The generative AI features of Copilot Studio transform how you build agents, significantly reducing manual work and configuration.

Prerequisites

Generative answers

With generative answers, your agent can find and present information from multiple sources, both internal and external. You don't need to manually create multiple topics that might not address all customer questions. Use generative answers as primary information sources or as a fallback when authored topics can't answer a user's query. By using generative answers, you can quickly create and deploy a functional agent.

What changed?

When a traditional chatbot can't determine a user's intent, it asks the user to rephrase their question. If after two prompts, the chatbot still can't determine the user's intent, it escalates to a live agent by using the Escalate system topic.

Before it escalates to a live agent, a Copilot Studio agent with generative answers capabilities uses natural language processing (NLP) to:

  • Parse what a user types to determine what they're asking.
  • Find, collate, and parse relevant information from specified sources, such as your company's website or multiple sources including SharePoint.
  • Summarize search results into plain language delivered to the agent user.

Your workflow might look like this:

  1. You create an agent and test it thoroughly. By default, your agent has generative orchestration turned on in its Generative AI settings.
  2. After testing, you publish your agent. It can instantly provide answers, help, and guidance to your customers.
  3. You create individual topics for frequently asked questions. These topics might develop from analytics from previous agents or existing support issues.

AI general knowledge

In addition to generative answers, your can let your agent use AI general knowledge to find and present information in response to your customer's questions. General knowledge saves you from needing to manually create multiple topics, which might not address all your customer's questions. It can also help when the existing topics can't address the customer's intent.

What is AI general knowledge?

AI general knowledge applies the capabilities of AI to access and provide information, insights, and assistance across a wide range of topics.

Why use it?

  • Accessibility: The agent can instantly access a vast repository of information and expertise across a wide range of subjects.
  • Versatility: It's capable of addressing diverse topics and tasks, making it a versatile resource for various needs.

Note

While AI general knowledge can provide valuable information and assistance, you need to critically evaluate the information it provides and consider consulting other sources for verification or further clarification when necessary.

What's supported?

AI-based authoring might be subject to usage limits or capacity throttling.

Note

Long prompts can cause generative answers nodes to fail due to usage limits or capacity throttling. Use short, focused prompts or split queries. Learn more in FAQ for generative answers.

Quotas

Quotas are default constraints applied to agents that limit how often you can send messages to an agent. The purpose of quotas is to throttle the client's service load, which protects a service from being overloaded and protects the client from unexpected resource usage.

Agents with generative answers capabilities have a limit on the number of queries they can make to derive answers from the URL you specified. Normal conversations that use agent topics follow the usual quotas and limitations.

Languages

For the list of supported languages, see Language support.

Related content

  • Last updated on