The rise of AI tools like ChatGPT has sparked understandable but unfounded concerns among content designers about the future of their craft. However, AI models are only as good as the instructions they receive. Since content designers are experts in communicating, they’re uniquely positioned to design and evaluate instructions for AI models to ensure high-quality outputs.
In this talk, we’ll see how content design played a critical role in recent AI text generation projects at Shopify. You’ll see how they defined a quality rubric for outputs, crafted prompt instructions, and built a system for identifying gaps to focus quality improvements.
In this talk, you’ll learn:
- The importance of identifying and describing the attributes of high-quality output. This allows you to create a clear rubric for evaluation, but also to measure your progress as you improve the prompts.
- Basic methods for iteratively crafting instructions that AI tooling will understand, including some lessons learned about how to troubleshoot and test prompts.
- Processes for measuring output quality and articulating quality gap progress to stakeholders as well as providing direction for iterative improvements.