Creating a Brand Kit

Lesson Overview

In this video, you'll learn about Brand Kits in AirOps, which are portable personas that provide important context to language models (LMs) to improve the quality of their outputs. You'll discover how to create, manage, and leverage Brand Kits across your AirOps workflows.

• 0:00: The importance of high-quality input for LMs
• 0:30: Introduction to Brand Kits and their purpose
• 1:07: Navigating to Brand Kits in the AirOps platform
• 1:55: Key fields in a Brand Kit (brand point of view, competitors, writing style, etc.)
• 3:36: Adding custom variables to Brand Kits

Key Concepts

Brand Kits as Portable Personas

Brand Kits are centrally managed personas that provide crucial information to LMs at runtime, enabling them to generate higher-quality outputs tailored to your brand's specific needs. These kits can be updated as your business evolves or as you identify gaps in the LM's knowledge, ensuring consistency across all workflows that utilize the Brand Kit.

The Importance of High-Quality Examples

The quality of the examples provided in a Brand Kit is a significant determinant of the LM's output quality. Instead of relying on instructional prompts, it's more effective to provide high-quality writing samples that demonstrate the desired tone, style, and characteristics you want the LM to emulate.

Leveraging Brand Kits in Workflows

Brand Kits can be used in their entirety or by referencing specific subvariables within a workflow. This flexibility allows you to be surgical in your approach, using only the necessary components of the Brand Kit for each specific task.

Key Takeaways

1. "The quality of the output is really dependent on the quality of the input." - Emphasizes the importance of providing high-quality information and content to LMs to ensure optimal results.
2. "Brand kits are portable personas that you can use across your Air Ops workflows that provide really important information to your LMs at runtime." - Defines Brand Kits and highlights their role in enhancing LM performance.
3. "Models do much better when you show versus tell." - Stresses the significance of providing high-quality example content in Brand Kits, as opposed to relying on instructional prompts, to achieve the desired tone and style in LM outputs.
4. "This is a really core point of leverage that will affect all of your downstream workflows that then go ahead and leverage this." - Underscores the centralized nature of Brand Kits and how updating them can positively impact all associated workflows.
5. "Rather than updating all of your prompts across 20 or 30 different workflows, go back to your brand kit, update it there, make sure that's kept up to date and is maintained well, and you're gonna get really high quality outputs and outputs that reflect the evolving nature of your business." - Summarizes the key benefits of using and maintaining Brand Kits in AirOps.