Updated: December 5, 2023

Design Principles for Generative AI

Intro

Generative AI can transform entire industries, and it’s already changing how we live and work. Its ability to generate novel content, whether text, image, data, or video, is a game changer. This technology is rapidly changing how we design, build and interact with software. 

As we race to bring the power of generative AI to our customers, we must do so intentionally, so they have a cohesive, high-quality experience across all SAP products. The generative AI design principles aim to enable and ensure this alignment as more and more teams incorporate this new technology by expressing a shared vision for what makes great generative AI experiences for our users. 

Approach

These guidelines and principles are customized to speak to the unique aspects and design considerations for generative AI. They have been created in alignment with other key SAP guidelines including: 

  • SAP Purpose & Values: Generative AI will have a tremendous impact on individual lives, broader society, and the environment. SAP’s sustainability mission to help the world run better and improve people’s lives is a north star that should guide us in choosing valuable generative AI use cases and ensuring that humans are always in the driver’s seat.
  • SAP’s Guiding Principles for AI, SAP’s Global AI Ethics Policy and our Guiding Principles for Designing Intelligent Systems: As a subset of AI, these principles also apply to generative AI. The generative AI guidelines are intended to provide guidance around the unique opportunities and considerations required when using this specific type of technology.

How generative AI design principles are tied to SAP's overall purpose, AI policies, standards, and generic principles.
How generative AI design principles are tied to SAP's overall purpose, AI policies, standards, and generic principles.

Generative AI Design Principles

We’ve defined five principles to inspire and guide you when designing generative AI features:

Each principle comes with concrete actions that you can use as criteria for evaluating your designs. 


Empower and Inspire

Use AI to enhance human capabilities and improve outcomes, not to replace human intelligence. Facilitate empowering and inspiring interactions with AI that will improve human lives.   

How we do this

  • We prioritize generative AI use cases that will bring the most value and differentiation to users and companies. 
  • We design interactions that foster collaboration and iteration with generative AI systems to amplify human intelligence, creativity, and curiosity. 
  • We frame generative AI results as inspiration, not fact. 
  • We proactively help users in verifying and improving generative AI results. 

Maintain Quality

Ensure high-quality input from the AI and the user to confidently derive the best possible results. Help users understand, analyze, and validate the output generated by AI. 

How we do this

  • We maintain contextual awareness of user needs and priorities. 
  • We provide education and guidance to users through embedded and interactive support. 
  • We enable users across diverse levels of literacy and technical expertise. 

Show the Work

Enable human evaluation of AI collaborations by capturing the user’s iterative journey towards the desired result.

How we do this

  • We identify AI-generated content. 
  • We cite the data sources informing generative AI results. 
  • We enable users to capture, save, edit, and share their prompt history and embed or summarize it into exported results. 

Continuously Assist

Assist users in avoiding errors. Educate them on the limitations of generative AI so they can spot and correct inaccurate results. 

How we do this

  • We clearly identify AI generated elements of content that require human validation. 
  • We provide progressively disclosed explainability behind AI generated outcomes. 
  • We enable the validation of results within the user experience. 

Humans Hold the Keys

Make it clear that people are in charge by identifying when and where AI is being used to generate novel content. 

How we do this

  • We explain how the AI works and the data it uses. 
  • We provide controls for customizing the AI’s parameters or turning AI enhancements off entirely. 
  • We clearly state the organization’s ethical criteria for using generative AI tools. 
  • We specify a user’s responsibility when using generative AI tools.