Getting started with the AI chat
The AI chat is the conversational experience in watsonx Orchestrate that enterprises can use to configure agentic behavior that intelligently chooses the best actions to meet the needs of users.
The chat is powered by an AI agent that gets context from users to better select that actions it should take, disposing of tools set for automation and AI assistants designed to help on specific topics, find information in internal documents, and guide the user through the process.
Before you begin
The following procedure provides basic high-level steps to work with the AI chat of watsonx Orchestrate. By following these steps, you get a better understanding on how to work with the AI chat and additional workflows you can use.
Prompting the chat for answers
You can ask questions to the AI chat and explore its answers that are built with large language models. By doing it, you get text-based answers that you can use for different purposes like: get more information on topics and format structured text. Go creative with the subjects that you query the AI agent.
Empowering the AI agent with AI assistants
By default, your AI agent can't run skills until you add AI assistants to it. This feature enables the AI agent to route your prompts and questions to an AI assistant that might be better to answer you. By adding AI assistants in your AI agent, you can also use AI assistants actions also to automate your work.
The following procedure describes the end-to-end workflow to empower the AI chat with AI assistants:
-
Identify skills in the catalog that you want to use to automate tasks
Identify the tasks of your work that you want to automate and find the skills that support automating them. If needed, you can have builders build more skills to better support your work.
On the skill catalog you can check the skills available for each skill set, the list of the available skills changes depending on the skill set that you selected.
You can use the personal skill set to browse the skills available in your tenant; the skill catalog has over 1000 prebuilt skills that you can use. The custom skills that are built by builders also appear there.
Some skill sets belongs to the AI assistants that you have in your tenant. To use skills on the AI chat, you can add them to one or more AI assistants skill sets, and then add these AI assistants to the AI chat. The managing of the skills available for each AI assistants skill sets is made by users with the administrator role on the Team skills page. Also, you can add skills to the AI chat through the AI agent.
-
Connect to the app of skills on the Orchestrate Agent skill set
Users with administrator role must connect to the apps in the Orchestrate Agent skill set by navigating to the Manage team page in the watsonx Orchestrate provisioned on AWS, or by navigating to the skill sets page in the watsonx Orchestrate provisioned on IBM Cloud. Administrators must connect with shared credentials.
-
Add apps and skills to the AI agent
Administrators add the skills of the connected apps to the AI agent so that these skills can be available in the AI chat. When a skill is being added, an accurate description of the purpose and capabilities of the skill is required for the AI model to successfully activate this skill in the chat. After adding the skills to the AI agent, administrators can define whether the app will be visible or not in the chat.
-
Optional: Build skills
It is possible that you and your team identifies that you need specific skills to automate particular tasks or process of your work and business, and the skills available do not match your requirements. Users of watsonx Orchestrate with builder role can add more skills to the skill catalog by building them on Skill studio.
Skill studio enables builders to create skills of various types, including: decision models, workflows, generative AI answers, connectors to application programming interfaces (API) through OpenAPI specifications, and flows that combine existing prebuilt or custom skills. With these skills, builders can automate tasks and processes with varying complexities.
Skills become available to users on the skill catalog after builders build, enhance, and publish them.
-
Create AI assistants
Use the AI assistant builder to create AI assistants. To create one, do the following steps:
-
Connect to the app of skills on the skill sets of the AI assistant
Each environment of an AI assistant is linked to a skill set. By default, AI assistants start with two environments: Draft and Live, thus, also starting with two skill setss; for example,
My_Assistant_Draft
andMy_Assistant_Live
. Users with administrator role must connect to the apps of all skills that builders want to add to AI assistants by navigating to the Team skill sets on the Manage team page or by navigating to the skill sets page if on a watsonx Orchestrate provisioned on IBM Cloud. Administrators must connect with shared credentials. -
Add actions to an AI assistant
Builders enhance AI assistant with actions to enable them to do complex activities, thus automating tasks and processes. Available actions include AI-guided actions, skill-based actions, or custom-built actions.
-
Publish AI assistants
When done, builders need to publish the AI assistants so that admins can add them to the AI chat.
-
Connect AI assistants to the AI chat
Admins must connect AI assistants to the AI chat so users can use their capabilities. Add AI assistants to the AI chat on the AI agent configuration page.
-
-
Prompt the AI chat
Now that the AI chat has one or more AI assistants, it can route your prompts to these AI assistants so they handle the work. These AI assistants run actions that can answer your prompt with text or guide you through the experience of automating tasks with skills.
You can prompt an AI assistant directly by starting your message with
@
Deepen your knowledge
See the main sections in the documentation to deepen your knowledge about how to use the AI chat.
-
Check out the prebuilt skills that are available by default on watsonx Orchestrate.
-
Learn how to add skills to the AI chat and see a guideline to describe the skill's capabilities.
-
Building skills in Skill studio
Learn how to use Skill studio to build powerful skills to automate your work. You can create projects that have decisions models, workflows, or generative AI, and publish them as skills for your team. Also, you can create skills from OpenAPI specification files, discover skills from other IBM or third-party products, or combine skills to build skill flows.
-
Building AI assistants in AI assistant builder
Learn how to create AI assistants on AI assistant builder to deploy engaging conversational experiences either on the AI chat of watsonx Orchestrate or embedded on third-party applications.
-
Learn more about how to configure and use the AI chat by exploring all of its components and details.