Using Orchestrate Chat

Edit online

The Orchestrate Chat page is your main workspace for interacting with AI agents, running tasks, and managing conversations. From here, you can start new chats, explore what each agent can do, and complete actions quickly - all with a simple, intuitive interface.

In IBM watsonx Orchestrate, AI agents help you automate tasks, coordinate workflows, and manage everyday processes. Agents can reply directly in the chat or work behind the scenes to complete actions and share information.

Users interact with agents through the chat interface, which serves as the entry point for all requests and responses. When a user submits a request, the system generates a response based on the agent’s configuration, including its prompts, instructions, and the selected Large Language Model (LLM).

The chat interface itself does not change how responses are generated. Instead, the LLM produces outputs according to the behavior and constraints defined in the agent configuration. This separation helps ensure that responses remain consistent with the agent’s intended design, regardless of the chat surface used.

Note:

AI-generated responses can vary. Always review and verify responses before acting on them.

Because LLMs are inherently non-deterministic, responses can vary slightly even when similar or identical inputs are provided. Differences might include wording, structure, or level of detail. These variations are expected and do not affect the accuracy, reliability, or intent of the information.

What Orchestrate chat can help you achieve

Orchestrate Chat connects you with AI agents designed to help you reach your goals, whether you are looking for answers, need help completing a task, or want to automate part of your workflow. Within the chat, you can:

  • Ask questions and receive answers, summaries, or insights

  • Request the agent to perform supported tasks that align with your objectives

  • Switch between agents based on the type of work you need to complete

  • Easily identify which agent responded using color‑coded avatars

    • Each agent uses a unique avatar color (such as blue, gray, or green) to distinguish their responses in the conversation

If your tenant has no active custom agents, watsonx Orchestrate automatically provides prebuilt fallback agents powered by large language models (LLMs). These agents help you complete essential tasks, including:

  • Answering questions

  • Summarizing content

  • Generating new content

  • Classifying text

  • Translating text

  • Analyzing sentiment

The fallback agents ensure that you can still make progress toward your goals even before your organization deploys custom agents.

Note:

  • A prebuilt agent appears only when your tenant has no active custom agents.

  • It is hidden from the agent selection menu once a custom agent is deployed.

  • Recent chat threads remain hidden until at least one custom agent is added.

Learn more about prebuilt agents.

Get started with Orchestrate Chat

Follow these steps to start a conversation with an AI agent in Orchestrate Chat:

  1. From the left panel, click New chat to start a new conversation.

  2. Select an agent from the Agents list. Each agent is designed for specific tasks.

  3. Type your question or task in the input box at the bottom of the page.

  4. Press Enter or click the send icon to submit your request. The agent responds directly in the chat.

Helpful controls and shortcuts

Use these controls and keyboard shortcuts to work more efficiently in Orchestrate Chat.

Category

Description

Manage chat threads

Select the overflow menu (⋮) to rename or delete a chat thread.

Format your message

For multi-line input, use: - Alt + Enter or Shift + Enter (Windows) - Shift + Return (Mac) These shortcuts enables you to format your message before sending it.

Change display language

Change the display language to see the UI, welcome messages, and starter prompts in your preferred language. Only languages enabled by your admin in Platform language settings appear in the list. To change the language: open the Language menu next to your profile icon and select a language.
Left-to-Right (LTR) and Right-to-Left (RTL) language handling

To ensure optimal readability for right‑to‑left (RTL) languages such as Arabic and Hebrew, the chat interface automatically adjusts message alignment based on the content being rendered. Any prompt or response containing at least one RTL character is displayed with right‑aligned layout. This behavior applies regardless of the platform or UI language. No user configuration or action is required.

In some cases, agent responses might be generated incrementally and contain a mix of left‑to‑right (LTR) and right‑to‑left (RTL) text. For example, a message might begin in English (LTR) and RTL characters such as Arabic appear later as the response continues to stream. When this occurs, the message layout might shift from LTR to RTL alignment during streaming. This is the expected behavior as the chat renderer dynamically adjusts alignment based on the detected script to maintain correct reading order and text flow.

Learn by watching

Explore how to use watsonx Orchestrate Chat to interact with agents and complete tasks.

Note: This video has no narration.

Using Chat to interact with agents

Transcript The video shows the user selecting the Employee Support agent and starting a conversation in the Orchestrate Chat, by asking to update their personal address. Then, the user changes to another agent that is named Compensation, by asking for their compensation benefits.

Known issues

For known issues related to Orchestrate Chat, see Known issues for using Orchestrate Chat.