What's new in IBM® watsonx Orchestrate®

Connecting to external apps

This release brings a streamlined approach for managing connections to external applications on watsonx Orchestrate, ensuring secure and efficient integration with the tools your team depends on.

Key features include:

• Authentication methods: Establish connections by using various authentication methods, such as API Keys, Basic Auth, Bearer Tokens, Key-Value Tokens, and OAuth2 (Authorization Code & Client Credential). These connections serve as secure bridges, enabling your agents to interact with external systems and services.

• Manage Connections page: From the Manage > Connections page, you can:

  • View all existing connections
  • Add new connections with custom IDs, display names, and authentication settings
  • Filter connections by environment, status, or authentication type
  • Edit connection details for both draft and live environments
  • Environments: Configure connections in both draft and live environments. This setup allows you to test the integrations before deploying them in production.

• Credential types: Choose between member credentials and team credentials to match your app's requirements. Each credential type supports specific parameters based on the selected authentication method.

This update empowers you with greater control, security, and flexibility when integrating external applications, making it easier to extend your agent's capabilities. Refer to Managing connections for more information.

Customizing agent styles

You can now choose the style for your AI agent directly within the Agent Builder UI. This setting determines how your agent understands and responds to user requests, selects appropriate actions, and engages with tools and agents to complete tasks. You can opt for either the default or ReAct style based on your wanted output quality, speed, and task complexity according to your use cases. For more information, see Choosing an agent style.

Adding IBM watsonx Orchestrate assistants as collaborator agents

You can now add assistants from IBM watsonx Orchestrate as collaborator agents directly from the Agent Builder UI. This feature helps you to create domain-specific agents that focus on particular industries or tasks. By enabling your agents to delegate tasks to these assistants, you can take advantage of their structured and deterministic logic. This setup allows you to reuse and repurpose existing assistants more effectively, improving task accuracy, and modularity.

To learn more, see Adding IBM watsonx Orchestrate assistants as collaborator agents.

Monitoring agent performance

Use the built-in performance monitoring tool in IBM watsonx Orchestrate to track how your agents and tools are working. This feature helps you quickly spot issues and improve how your agents respond to user requests.

Key features include:

All agents at a glance

Get a high-level view of how all your agents are performing:

• Total messages: See how many messages all agents processed, including successful and failed ones.
• Failed messages: Track errors like service connection failures or server issues.
• Latency average: Measure how long, on average, it takes agents to respond.

Individual agent metrics

Drill down into each agent’s activity:

• View the agent’s name, description, number of requests, failed requests, and average latency.

Agent message traces

Understand how each message was handled:

• See the timestamp, trace ID, status, model used, and latency for every message.

Detailed trace view

Investigate specific issues:

• Review detailed error messages to help you fix problems in the agent’s setup or tool connections.

This feature empowers you to quickly detect and fix issues, improve response times, helping ensure agents run smoothly and reliably. For more information, see Agent analytics.

Importing tools from MCP servers

You can now integrate IBM watsonx Orchestrate with Model Context Protocol (MCP) servers, making it easier to import external tools and expand your agents’ capabilities.

MCP is a standard that allows secure and flexible communication between AI agents and external tools or data sources. With this integration, you no longer need to build custom APIs or manage manual configurations saving you time and effort.

After you connect to the servers, you can import tools and add them to your agents directly from the MCP tools list. Use filters to quickly locate the required tools and set the "Added to agent" switch for the tool to "Yes".

Manage MCP connections by editing connection details, removing unused servers, and monitoring all imported tools from the Connections page.

This update allows you to instantly enhance your agents capabilities by sharing tools across platforms. To learn more, see Importing tools from an MCP server.

Uploading files into chat threads for context-aware responses

You can upload documents directly into chat threads, enabling agents to generate responses based on the content of your files.

Supported file formats include .txt, .pdf, .csv, .docx, .xlsx, .pptx, and .htmlx. After you upload a file, the system validates the file and also provide clear error messages for any issues, such as failed loading, exceeded upload limits, or invalid files.

This feature helps ensure your agents respond with precision and context, by using the documents that you provide, making your conversations more productive and informed. For more information on how to upload a file in the chat, see Uploading files during a conversation.

Mapping data in flows

You can now enjoy enhanced data-mapping options that make configuring data flows between nodes easier and more efficient. With automapping, you can achieve seamless transfers effortlessly. If you prefer more control, you can customize parameters by using blank values, Python expressions, variables, literal values, or defaults tailored to your flow's needs.

These improvements give you greater control and precision over node input parameters, ensuring that your flows are exactly how you want them. Learn more about using data mapping in flows by checking out Data mapping.

Adding custom logic with code blocks

You can now use code block nodes to run Python code directly within your flow. These nodes enable you to insert custom logic as Python code, providing flexibility for data transformation, complex logic implementation, or message formatting.

Code block nodes are ideal for scenarios where standard nodes cannot fulfill your data processing, intricate logic, or message formatting requirements. Access flow variables, write Python by using built-in libraries like datetime, math, or JSON, and pass outputs to the next node.

For more information on how to use code blocks to create flows, see Code block examples.

Uploading and downloading documents in tools

You can now upload and download documents when using tools in both AI agent chat (where you interact with a deployed agent to accomplish tasks) and preview chat (for testing your agent during development).

• Upload files into tools that are set up to receive them.
• Download files generated by tools during a session.

Builders can configure these tools by using the watsonx Orchestrate Agent Development Kit (ADK) to support file uploads and downloads. This gives you a smoother experience when working with document-based tools, and allowing builders to control how files are handled.

Selecting a model for your AI agent

You can choose the foundation model for your AI agent directly within the agent builder UI. This feature provides greater control and flexibility in customizing your AI agent's capabilities. To learn more, see Choosing a foundation model.

Connecting to a content repository on a custom service

You can now connect to a content repository on a custom service, allowing you to use your existing custom service instances, whether hosted on the server side or client side. This enhances your capabilities by integrating your search functionality with a server to deliver search results.

This feature provides you with greater flexibility and customization options for integrating your custom services with watsonx Orchestrate, enabling access to and use of content from your preferred repositories.

For more information, see Connecting to a content repository on a custom service.

New prebuilt agents are available

watsonx Orchestrate now offers the following prebuilt agents.

Procurement

• Sourcing and contracts (Coupa)
• Supplier management (SAP Ariba)
• Invoice management (Coupa)
• Catalog search and supplier search (Coupa)

HR

• Learning and development (Oracle HCM)

Productivity

• Issue and bug tracking (Jira)
• Project and resource management (Adobe Workfront)
• Storage and collaborate file (Google Drive)
• Communication and messaging management (Slack)
• Account information (Salesforce)

For more information on the newly introduced prebuilt agents, see List of prebuilt agents.

Preview - Connecting to channels

You can now connect your agents to communication channels such as WhatsApp (through Twilio) and Microsoft Teams, making it easier for your team to interact with agents directly from the tools they already use.

For more information, see Connecting to channels.

Preview - Configuring voice

You can now enable voice interactions, allowing for verbal communication with users. On watsonx Orchestrate, you can configure voices that are powered by IBM Watson Speech to Text and Text to Speech services.

Key benefits include:

• Creating voice configurations
• Managing voice configurations
• Enabling voice modality

With this feature, you can create more dynamic and human-like interactions, making your agents more engaging and accessible. For more information, see Configuring a voice (Preview).

Other enhancements and changes

The following table details the enhancements that are delivered in the current release corresponding to the user role:

IBM® watsonx Orchestrate® API updates

See the updates in the IBM watsonx Orchestrate API for this release.

Documentation updates

See the updates in the watsonx Orchestrate documentation and other contents for this release.

New sections

• The Connecting to Microsoft Teams and Connecting to WhatsApp with Twilio sections was added to provide detailed steps on how to connect to the respective apps.

• The Evaluating an agents performance section explains how to use the performance monitoring tool to track message volume, latency, and failure rates.

• The Configuring a voice (Preview) section describes how to set up voice capabilities for agents by using IBM Watson Speech services.

• The Managing connections section details how to create and manage secure connections between external applications and watsonx Orchestrate.

• The Connecting to a custom service repository section was added to provide details on how to connect an agent to a custom service repository.

• The Importing tools from an MCP server were added give detailed information about how to connect to MCP servers and import external tools into your agent.

• The Choosing an agent style section provides information about agent styles, and how to set up agent style for your agent.

• The Agent creation specifications section gives you a high-level understanding of the fundamental components that form an AI agent.

• The Adding external A2A agents using ADK section provides information about adding external A2A agents as collaborators.

• The Connecting to apps section provides detailed information about how to connect to each application using the UI.

Updated sections

• The Known issues and limitations content was restructured to separate known issues for skill-based and agentic experiences.

• The Technology preview label was removed from features that are now generally available in this release.

• The Known issues and limitations topic now has the following entry:

  • Importing watsonx Assistants
  • Document upload fails
  • Repetitive responses by model Llama 3.2 70B Instruct
  • Llama 3.2 70B Instruct model context length
  • Chat with documents is not supported in voice (Preview)
  • Embedded chat disappears

New data center available on AWS

The AWS data center is now expanded to include the following region:

Mumbai

IBM watsonx Orchestrate data center now expands to Mumbai, enhancing global accessibility, performance, and adherence to local data sovereignty regulations. This expansion guarantees a scalable and dependable infrastructure, customized to cater to the specific requirements of users, while fostering business growth and providing an exceptional user experience throughout the region.

For more information, see Outbound IP addresses.

Support for new plans on IBM Cloud

Introducing the Premium plan

IBM watsonx Orchestrate introduced the Premium plan, by complementing the Essentials and Standard plans. This plan is designed for enterprises and requires advanced features and higher resource allocations. See what you get with the Premium plan:

• 1 million skill runs per month

• 50,000 monthly active users (MAUs)

• Data isolation capabilities: The data isolation feature allows organizations to create virtual data boundaries within a multi-tenant environment, ensuring enhanced data security and compliance.

• Also, you get everything included in the Standard plan, which includes automations for workflows, AI decisions, document processing, and access to Skill Studio and Agent Builder.

For more information, see IBM watsonx Orchestrate on IBM Cloud Premium.

New plans for the agentic experience

IBM watsonx Orchestrate is introducing 6 new plans by allowing customers to fully use the agentic experience. This includes:

• IBM watsonx Orchestrate on IBM Cloud Essentials Agentic monthly active users
• IBM watsonx Orchestrate on IBM Cloud Standard Agentic monthly active users
• IBM watsonx Orchestrate on IBM Cloud Premium Agentic monthly active users
• IBM watsonx Orchestrate on IBM Cloud Essentials Agentic Messages
• IBM watsonx Orchestrate on IBM Cloud Standard Agentic Messages
• IBM watsonx Orchestrate on IBM Cloud Premium Agentic Messages
• IBM watsonx Orchestrate on IBM Cloud Trial

These plans depend on two usage models:

Agentic Monthly Active Users (MAUs): Tracks the number of unique monthly active users interacting with the AI agents.

Agentic Messages: Measures the volume of messages exchanged with AI agent.

Each agentic plan also includes access to the Agent builder that provides a low-code, self-service experience to build AI assistants and custom agents. These options allow you to choose the best one that suits your business needs.

For more information, see Licenses and entitlements for watsonx Orchestrate on IBM Cloud.

Customizing with Add-on options

To further customize your experience, watsonx Orchestrate on IBM Cloud offers additional services:

• IBM watsonx Orchestrate on IBM Cloud Add-on Skill Runs
• IBM watsonx Orchestrate on IBM Cloud Add-on monthly active users
• IBM watsonx Orchestrate on IBM Cloud Class Inference
• IBM watsonx Orchestrate Add-on Voice monthly active users
• IBM watsonx Orchestrate Add-on Voice Minutes
• IBM watsonx Orchestrate Add-on Messages
• IBM watsonx Orchestrate Add-on Documents
• IBM watsonx Orchestrate Add-on Domain Agents
• Non-IBM products Add-on Domain Agents

For a detailed comparison of the new plans and to determine which edition best suits your organization's needs, see Licenses and entitlements for watsonx Orchestrate on IBM Cloud.

Data isolation for enhanced security

watsonx Orchestrate on IBM Cloud now supports data isolation, a key feature for organizations with strict security and compliance requirements, including HIPAA.

With data isolation, each customer gets a dedicated data storage instance, ensuring their information is fully separated from other tenants. This feature helps protect sensitive data and gives you greater control over access and usage.

Data isolation is available with the Premium plan on IBM Cloud.

This enhancement strengthens data security and helps your businesses meet regulatory standards. For more information, see Data isolation on IBM watsonx Orchestrate.

Other enhancements and changes

The following table details the fixes that are delivered in the current release corresponding to the user role:

Documentation updates

See the updates in the watsonx Orchestrate documentation and other contents for this release.

New sections

• The Data isolation on IBM watsonx Orchestrate section explains how watsonx Orchestrate isolates data on IBM Cloud.

Capturing user feedback

You can now provide feedback on agent responses by using Like and Dislike buttons. This new feature makes it easier to evaluate and enhance agent performance based on real user input. The builder can view feedbacks that were submitted by users and use the feedback to gather insights and continuously improve the agent experience. This feature includes:

• Like or Dislike buttons to let users quickly rate responses.
• Options to select feedback reasons and add optional comments.

Feedback is stored with each message, including the selected reasons and an extra description, when it is provided. For more information, see Capturing user feedback.

Support for agentic plans on AWS

IBM has introduced a new tiered structure for watsonx Orchestrate on AWS, comprising three distinct editions:

IBM watsonx Orchestrate Agentic Essential:

Designed for small teams or those starting with AI automation. You get 600 Resource Units (RUs) and support for up to 40,000 Monthly Active Users (MAUs). It includes access to Skill Studio, where you can build and deploy AI-powered workflows, decision services, and document processing.

IBM watsonx Orchestrate Agentic Standard:

Suitable for growing businesses, with this plan you get 6,000 RUs per month. It includes everything in Agentic Essentials, and more capacity to scale your digital workforce.

IBM watsonx Orchestrate Agentic Premium:

Tailored for enterprises requiring advanced capabilities and higher resource allocations. This plan provides you with 7,500 RUs monthly, and everything in Agentic Standard plan.

This expansion provides businesses with more tailored options to match their specific requirements.

Customizing with Add-on options

To further customize your experience, subscribe to optional add-ons, which includes:

IBM watsonx Orchestrate Add-on Documents:

It is an optional subscription for Essentials, Standard and Premium editions that provides additional document entitlements in packs of 1,000.

IBM watsonx Orchestrate Add-on Domain Agents:

It is an optional subscription for Essentials, Standard and Premium editions that provides you with prebuilt, ready to use domain agents.

These add-ons provide the flexibility to scale and adapt the platform to your business needs.

The introduction of the Premium plan and the enhanced resource allocations ensure a seamless provisioning and a consistent functionality across all components, allowing customers to fully use the agentic experience. For a detailed comparison of the new plans and to determine which edition best suits your organization's needs, visit the Licenses and entitlements for watsonx Orchestrate on Amazon Web Services.

IBM® watsonx Orchestrate® API updates

See the updates in the IBM watsonx Orchestrate API for this release.

Capturing user feedback API

You can now provide feedback on agent responses by using "Like" and "Dislike" icons. The parameter message_state stores the feedback details for a specific message in the thread. For more information, see Update a message.

Other enhancements and changes

The following table details the fixes that are delivered in the current release corresponding to the user role:

Documentation updates

See the updates in the watsonx Orchestrate documentation and other contents for this release.

New sections

• The Capturing user feedback ratings section was added to provide detailed information on the feedback collection for agent responses in chat.

Updated sections

• The On-premises offering section was updated to add information about how to use different endpoint URLs depending on the resource, which include methods from the IBM watsonx Orchestrate API and AI assistant builder API.

• The Licenses and entitlements for watsonx Orchestrate section is updated to include the new agentic plans available on AWS.

• The Known issues and limitations topic now has the following entry:

  • An error while running tools

Building flows from Agent builder

You can now create flows from the Agent builder. These flows help you to design agents to complete tasks.

Creating flows from the Agent builder enables you to visually connect tools and define logic in the Tool builder screen. You can add tools, set conditional branches with Python expressions, and define inputs and outputs without writing any code.

It is a faster and more intuitive way to create tools for agents, which adds more flexibility to the way the agents work. For detailed steps to create flows, see Creating flows.

Prebuilt agents for team messaging and collaboration

IBM watsonx Orchestrate now includes a new set of prebuilt agents designed to streamline communication and collaboration in Microsoft Teams.

The Team Messaging and collaboration is the manager agent, which intelligently routes your requests to the most relevant supporting agents. These agents handle a wide range of Microsoft Teams tasks, which helps to automate everyday collaboration activities. To learn more, see Productivity agents

Connecting prebuilt agents to third-party apps

You can connect prebuilt agents to external applications using CLI commands from the Agent Development Kit. This integration process makes it easier to establish interactions between agents and third-party applications.

After you configure the connection, the agents can perform their tasks more efficiently, as they can leverage the functionalities of the external applications directly. For more information on how to connect prebuilt agents to SAP Successfactors and Workday, see Connecting to apps.

Using document classifiers in Skill studio (Beta)

You can now use an AI-powered document classifier to identify specific document types, such as invoices or contracts, and automatically classify uploaded documents. For more information, see Creating document classifiers.

Language support for document extractors in Skill studio

You can now set the language of documents when configuring a document extractor. Available options are English, French, or English with handwriting.

The English with handwriting allows you to extract a mix of typed text and handwritten text from your documents in English. This feature helps you efficiently process documents that require manual input, such as insurance policies or purchase orders.

Agent name and Avatar now visible in chat

You can now see the agent's name and avatar when using AI agent chat, preview, or embedded chat. This update makes it easier to identify the agent you are interacting with, ensuring a more personalized experience, even when switching between agents.

Other enhancements and changes

The following table details the fixes that are delivered in the current release corresponding to the user role:

Documentation updates

See the updates in the watsonx Orchestrate documentation and other contents for this release.

New sections

• The Team messaging and collaboration agent and its associated agents are added to the productivity agents list.

• The Getting started with the watsonx Orchestrate section provides detailed information on how to start using agents on watsonx Orchestrate.

Updated section

• The documentation topics were updated to emphasize the agentic experience.

• The Monitoring skills and Prebuilt charts section now include a deprecated label.

• The Switching from agentic experience to skills experience content is updated to rearrange the steps for switching to the skill based experience.

• The Known issues and limitations topic now has the following entry:

  • Authentication on watsonx Orchestrate API with IP address restriction
  • Enhancing an existing skill