Using AI Tool Builder

Predefined AI tools can generate social media posts, blog posts, and product banner text for common use cases. However, your business may have unique requirements that need specialized prompts or specific data grounding.

The AI Tool Builder is an administrative feature within the SAP CX AI Toolkit that allows you to:

• Validate prompts grounded in your own SAP Customer Experience data.
• Create custom AI tools tailored to your specific business processes.
• Activate or deactivate tools to control what is available to your business users.

For SAP Commerce Cloud users, this means you can build tools that directly reference your product catalogs to automate the creation of marketing copy, technical specifications, or category descriptions.

When creating an AI tool, you define how the AI should behave and which data it should "see" to produce accurate results.

To start, navigate to the AI Tool Builder in the left panel and select Create AI Tool.

You will need to provide:

1. Identity: A name, short description (for the tool card), and long description (for user instructions).
2. Licenses: Licenses that can use the tool once it is activated.
3. Source Application: For this context, select SAP Commerce Cloud. This allows the tool to access your product data.
4. Prompt: Create the instruction prompt that guides the AI in generating responses based on referenced data (grounding).
5. Default Tone and Length: Set the preferred "voice" (for example, professional, enthusiastic) and output size.
6. Special Instructions: Enable this to allow users to add extra personalization (for example, "Make it sound like a summer sale") when they run the tool.

Grounding is the process of providing the AI with business context. In the Tool Builder, you use specific placeholders to tell the AI which data to retrieve from your catalog.

When the Source Application is set to SAP Commerce Cloud, you can reference two primary entities:

• Product Entity: Used for tools focusing on individual items.
  • {{Product.name}}: Inserts the name of the product.
  • {{Product.*}}: Pulls all available attributes for a specific product.
• Category Entity: Used for tools that need to understand the broader grouping of products.
  • {{Category.name}}: References the name of the selected category.
  • {{Category.List.Product}}: This is a merge field reference. It provides the AI with access to data from multiple related products within the selected category. This is essential for creating content that summarizes a collection or identifies common trends across an entire product line.

The following video demonstrates how to build a custom tool named the Category Page Content Creator. It is designed to look at a list of products within a category and generate an engaging category landing page content that highlights common themes and features.

Key configuration highlights:

• Tool Setup: The administrator names the tool and sets the source to SAP Commerce Cloud.
• The Prompt Construction: The administrator writes a prompt that leverages the data context field to ensure the AI reads the product list.
• The Result: When a user runs this tool on a specified category, the AI uses the data context field to analyze the selected products in that category and generates a description that accurately reflects the specific assortment.

To ensure your custom AI tools produce high-quality results, follow these best practices:

• Define the Role: Tell the AI who it is (for example, "You are an expert e-commerce Copywriter for a high-end fashion brand").
• Define the Outcome: Be specific about what you want (for example, "Generate a 3-paragraph summary of these products").
• Choose Zero-Shot or Few-Shot Prompt accordingly:
  • Zero-shot: No examples provided; the AI relies on its base training.
  • Few-shot: You provide one or two examples of "Good" output within the prompt to guide the AI's style.
• Avoid Sensitive Info: Never enter personally identifiable information (PII) into the prompt field.