Comprehensive Guide to start with IBM watxsonx Assistant

Comprehensive Guide to Start with IBM watsonx Assistant

IBM watsonx Assistant is a next-generation conversational AI solution that empowers individuals and organizations to build generative AI assistants effortlessly. These assistants deliver seamless self-service experiences to customers across various devices or channels, boost employee productivity, and scale efficiently across a business.

In this guide, you will learn how to use watsonx Assistant to build and deploy an intelligent conversational AI assistant solution that simplifies tasks, improves business efficiency, and amplifies productivity.

Before You Begin

Follow the instructions to access the IBM Cloud account provisioned to be used in this hackathon.

Each participant is provisioned with an individual IBM Cloud account for the hackathon. Since teammates cannot be added to your account, you’ll need to collaborate offline or virtually, working on one team member’s assigned IBM Cloud account.

Plan Limitations

The hackathon-provided IBM Cloud account is pre-configured with a watsonx Assistant Trial plan and includes necessary services to complete the hackathon. However, note the following limitations:

• No Support for:
  • Built-in conversational search
  • V2 API endpoints

Refer to watsonx Assistant Trial plan limits.

IBM has also added a $100 credit on top of the free tier for pre-configured services. You’ll receive email alerts at 25%, 50%, and 80% of your credit use, with an account deactivation notice within an hour upon reaching 100%. Plan your usage wisely and back up your work regularly. Refer IBM cloud usage tips and saving your work.

Important: Notifications are only sent when limits are reached hourly. Exhausting the credit limit within an hour may result in deactivation without a warning email.

Usage Restrictions

• Participants are not allowed to use the provisioned hackathon IBM Cloud account services for any purposes other than to build a solution for the hackathon.
• To use any services other than the ones pre-configured in the hackathon provisioned cloud account, you will have to bring your own personal IBM Cloud account and utilize the services.
• If you notice a permission/access issue for any other services on the provisioned cloud account, then they are not required/available for this hackathon.

Data Sets

As you collect data for your project, it’s important to follow best practices. Here are some helpful tips:

• You are responsible for ensuring data compliance.
• Public website data may be used if the terms allow for commercial use, but make sure to keep a list of the websites you reference.
• Do not use data or assets containing company confidential information, or any other data without the permission of the data owner. Ensure you have proper approval.
• Do not use any client data.
• Do not use any data containing personal information (PI).
• Do not use data obtained from social media.

Step 1: Setting Up Your IBM Cloud Account

To ensure a smooth onboarding process, begin by setting up your IBM Cloud account following these steps:

1. Access Your IBM Cloud Account:

   • Check the email you used for hackathon registration for an invitation from the IBM Cloud team.

   • If you can’t find the email, search for “IBM Cloud” in your inbox or check your junk/spam folders.

   • Click the Join Now button in the email.

     

2. Create or Log Into Your Account:

• If it’s your first time setting up an IBM Cloud account:

  • Enter a new password, complete your personal information, accept the terms, and click Continue.

    

  • Acknowledge the creation of your IBMid and click Join Account.

    

• For existing IBM Cloud accounts:

  • Simply read and accept the terms and click Join Account.

    

3. Complete Authentication:
   • Click Continue to finish the authentication and gain access to the IBM Cloud dashboard.

     

Note: If you have an existing account, you might be directed to your current dashboard. Ensure you switch to the designated WatsonX Challenge or Watsonx Ecosystem Challenge account.

Step 2: Getting Started with IBM watsonx Assistant

Now that your IBM Cloud account is set up, follow the steps below to create your first assistant and get started with building your AI solution.

1. Accessing the Service:

   • Navigate to the Resource list on your IBM Cloud dashboard.

     

   • Expand the AI/Machine Learning section and select the pre-configured watsonx-Challenge WA service.

   • Click Launch watsonx Assistant.

     

2. Creating Your First Assistant:

   • Enter a name and optional description.

   • Set the language to English and click Next.

     

   • Personalize your assistant with display settings such as colors and avatars.

   • Review your setup and click Create.

     

   • An Assistant will be created, and you will be taken to its home page. Now you are ready to begin building your first assistant.

     

Step 3: Planning Your Assistant

With your initial setup complete, it’s time to think strategically. Planning your assistant will help you build a solution that aligns with your goals and user needs.To better understand how to effectively plan your assistant, refer to this guide.

Proper planning is essential to ensure that your assistant is well-structured and meets its intended goals. Here’s how to start: Before building your assistant, plan out:

• Initial Channel Selection: Decide which platform or channel your assistant will operate on.
• Domain Expertise: Choose the primary focus (e.g., customer support, HR).
• Tone and Language: Set the communication style.
• Content Source Integration: Identify relevant content sources.
• Handoff Strategy: Plan how interactions will transfer to human agents if needed.

Step 4: Building Actions and Steps

Once you have a plan in place, it’s time to start constructing your assistant by creating actions and steps. These form the foundation of your assistant’s interactions.

Creating actions and steps is critical to constructing a robust conversation flow. Follow these guidelines to build effectively:

• What is an Action?: A task or problem the assistant helps resolve, like paying a bill or tracking orders.

  

• What is a Step?: An interaction between the assistant and user, including clarification questions, final responses, and agent handoffs.

Example Exercise:

This section includes two example exercises: Building a Billing Use Case and Travel Assistant Use Case Exercise: Integrate watsonx.ai LLM for Gen AI, each designed to help you build essential components and explore generative AI integration within your assistant.

Building a Billing Use Case

1. Create Your First Action:

   • On your assistant's home page, select Actions from the left navigation panel.

     

   • Click the Create action + button and choose Start from scratch.

     

     

   • Train the assistant by entering example phrases such as “I want to pay my cable bill please” and save them to help the assistant recognize related topics.

     

2. Add the First Conversation Step:

   • Use the pre-created blank step to add a clarification question in the Assistant says text box, like “What type of account are we talking here?”.

     

   • Define customer response options (e.g., Cable, Internet, Phone) and apply the changes.

     

     

3. Add Conditional Steps:

   • Create a new step asking for the account number only if the selected option is “Cable”.

     

     

   • Set conditions so that this step only appears when Step 1's answer is “Cable”.

     

4. Agent Handoff Step:

   • Add a step saying, “Let me get you to an agent who can help you pay your cable bill!” and set it to connect to an agent under specific conditions (e.g., when Step 1 is “Cable”).

     

   • Configure the message to the agent to include: “This user wants to pay their cable bill. Account number is $[variable from Step 2]”.

     

5. Handling Internet and Phone Bill Requests:

   • Create a step with text like, “To pay your bill, you can head to our online portal.”

     

   • Insert a variable for from Step 1 and link to your billing portal.

     

     

   • Apply conditions to ensure this step triggers when the answer is “Internet” or “Phone”.

     

     

• Finally, under And then section, select Continue to next step drop down and select End the action option.

  

6. Testing the Flow:

   • Preview the conversation flow by using the Preview button and simulate different scenarios (e.g., paying a cable bill, selecting Internet).

     

   • Add more training examples if the assistant does not recognize initial phrases effectively.

     

Travel Assistant Use Case Exercise: Integrate watsonx.ai LLM for Gen AI

Try the Travel Assistant Use Case Exercise to integrate watsonx.ai LLM and enable generative AI capabilities in your assistant.

Follow the watsonx.ai integration exercise guide to build a Travel Assistant conversation that incorporates watsonx.ai LLM. This exercise will introduce you to integrating watsonx.ai AI models into your assistant, though it is not intended as a final submission for the hackathon.

Step 5: Visualizing and Reviewing Your Assistant

After building your actions, reviewing and visualizing them is crucial. This helps ensure your workflow is logical and meets your project requirements. For more comprehensive details on building and refining your assistant, refer to this documentation.

A visual representation of your assistant’s actions helps in refining and communicating its logic. Here’s how to navigate this feature:

• Toggle between Edit and Visualization tabs to see action flows as a flowchart.
• Use this visualization for better understanding and presentation.

  

Step 6: Previewing and Customizing

Before final deployment, previewing your assistant will help identify areas for improvement and ensure that it aligns with your brand. Customize as needed to perfect the user experience.

Previewing your assistant allows you to test its functionality and make necessary adjustments. Customize it to align with your brand using these steps:

• Preview Your Assistant:
  • Test functionalities on the Preview page.
  • Use options like Copy link to share or Change background to simulate different user experiences.
• Customize the web chat to align with your brand.For more information on configuring web chat, refer to this resource.

Step 7: Publishing and Deploying

With your assistant ready and tested, the next step is to publish and deploy it. This ensures your assistant is accessible to your users in the desired channels.

Once your assistant is built, publishing and deploying it ensures it reaches your target audience. Follow these procedures to manage deployment:

• Publishing:
  • Maintain version control to manage content updates and rollbacks. For more information on publishing, refer to this guide.
• Deployment Options:
  • Deploy on websites, mobile apps, or integrate with phone systems and other channels.

Step 8: Analyzing Your Assistant

After deployment, continuous analysis is key to maintaining and improving your assistant. This section will guide you on how to effectively monitor and optimize its performance.

To continuously improve your assistant, analyzing its interactions is vital. Use these insights to enhance user satisfaction and performance: Use the Analyze page to track interactions, understand user needs, and identify areas for improvement. For more details on analyzing your assistant, refer to this resource.

Save your work

Make sure to save any work if you want to retain for your records. The provisioned IBM Cloud accounts will be disabled anytime after the end of the hackathon. Follow the Backing up and restoring data documentation to save your work.

Conclusion

In this comprehensive guide, you have learned how to set up your IBM Cloud account, create your first watsonx Assistant, plan and build effective actions and steps, visualize and review your work, and finally, publish, deploy, and analyze your assistant. By following these steps, you can build a robust and responsive assistant tailored to your needs. Remember, continual refinement and analysis will help you maintain a high-quality user experience.

With these insights, you are now well-equipped to harness the full potential of IBM watsonx Assistant and create impactful AI solutions. Good luck with your development journey!

Additional Resources

To expand your understanding and capabilities with watsonx Assistant, explore these additional resources and tools.

For further learning and deeper customization, explore these additional resources to boost your assistant’s capabilities:

• Building actions from templates
• Starting the conversation
• Understanding your users’ questions or requests
• Boost words
• Adding assistant responses
• Collecting information from your customers
• Using variables to manage conversation information
• Adding conditions to an action
• Adding conditions to a step
• Choosing what to do at the end of a step
• Calling a custom extension
• Saving your actions
• Reviewing and debugging actions
• Global settings for actions

watsonx Assistant Trial plan limitations

• Does not support:
  • Built-in conversational search
  • V2 API endpoints
• Actions:
  • 10 assistants per instance
  • Up to 100 actions
  • 5 custom extensions per assistant
  • Session inactivity timeout 5 minutes

IBM Cloud account usage tips

General:

• Collaborate with other participants to form a team and leverage your team members’ cloud accounts to utilize the resources effectively.
• Always save your work in regular intervals.

watsonx.ai: If you have integrated watsonx.ai to enable generative AI capabilities for your watsonx Assistant, the watsonx.ai service consumes resources by tokens for using the foundational models available on the platform. A token is a collection of characters that has semantic meaning for a model. Tokenization is the process of converting the words in your prompt into tokens. You can learn more about tokens and how they are calculated here.

Leverage the model parameters to set limits for token consumption and repetitions for efficient prompting. You can learn how to work with model parameters here.

Note: The focus of this hackathon is watsonx Assistant and NOT watsonx.ai. watsonx.ai must be used only for enabling the generative AI capabilities for your assistant and not for any other purpose. Your cloud account will be deactivated if any suspicious watsonx.ai service utilization is noticed.