How to make clear instructions for value generation

This guide gives some tips and tricks on how to provide instructions when working with AI value generation in Bizzkit PIM to ensure high-quality, consistent, and scalable product content. AI value generation is most effective when instructions are clearly structured across three levels:

• Category level: Defines tone of voice, category audience (if differentiating across the webshop) and communication strategy
• Attribute level: Defines structure and formatting
• Custom instructions: Used for one-off adjustments

Overview

To get the best results from AI text generation, it is important to:

• Separate what you say from how you say it
• Keep configurations consistent and non-conflicting
• Provide strong example data (best practice products) that supports instructions

Category Level Configuration

The category level acts as the foundation for all generated content inside a category.

This is where you define:

• How your brand communicates within a product category
• Who the target audience is, if it differs from one category from another
• What should be emphasized in the messaging

What to include

Tone of voice

Define how the text should sound:

• Professional or conversational
• Technical or simple
• Inspirational or sales-driven
• Etc.

Target audience

Specify who the content is written for:

• Professionals vs. beginners
• Younger vs. older audiences
• Men vs. women
• Etc.

Communication focus

Clarify what matters most in this category:

• Product specifications
• Design and aesthetics
• Price and value
• Durability and quality
• Etc.

Golden products (Best practice examples)

Select 2–5 high-quality products that demonstrate:

• Correct tone of voice
• Strong structure
• Appropriate level of detail

These examples are critical, as they guide the AI in generating consistent and relevant output.

Best practice

• DO: Use category level for meaning, tone and direction
• DON’T: Do not define structure (length, formatting) here

Attribute Level Configuration

Attribute-level instructions define how the generated text should be structured, formatted, and any specific constraints to consider.

Note that it is not always necessary to configure every attribute. The Bizzkit AI Assistant leverages best-practice examples to infer structure and formatting, especially for attributes like long descriptions, where existing product patterns provide strong guidance. In such cases, explicit instructions may be redundant.

What to include

Length

• E.g. 2–3 sentences
• E.g. Maximum 100 words

Structure

Define how the text should be built e.g.:

• Start with product name
• Describe usage
• Highlight key feature
• Call-to-action

Formatting

Specify output format where relevant:

• Plain text
• Bullet points
• Bold

Best practice

• DO: Use attribute level for structure and output format
• DON’T: Do not redefine tone of voice here

Custom Instructions

Use for one-off adjustments

Custom instructions allow you to temporarily adjust generation output.

These instructions are applied to specific attribute value and are not part of the permanent configuration.

When to use

• Campaign-specific content
• Quick experiments

Best practice

• DO: Use for temporary or situational changes
• DON’T: Do not rely on custom instructions for reusable logic

Final recommendations

To achieve the best results:

• Invest time in strong category configurations
• Use high-quality example products
• Keep attribute instructions clear and consistent
• Use custom instructions sparingly and intentionally
• Regularly review and refine configurations