Creating generative AI features on the web has become easier using Open AI's APIs.

Designers and developers can connect the API to websites and products, and provide in writing what they want the AI to know, do, and how it should behave.

The AI will give human-friendly answers to questions based on its knowledge. It can also pull knowledge from data sources and perform actions on behalf of users when provided with custom API endpoints.

To further the experience, crafting a brand personality brings it all together. When someone interacts with the AI, it should feel appropriate. Its tone of voice, design, and interactivity should all fit together to bring the best experience to users.

Making it trustworthy

Businesses often worry that the AI could say something inappropriate.

We've discovered that the API controls to avoid sensitive topics are solid enough to trust, making it highly unlikely that the AI says something inappropriate and unaligned with the brand's values.

A more significant concern with AI is hallucinations – instances where AI confidently says things when it doesn't know the exact answer. It means the AI can say stupid things while letting the user think it knows what it's talking about. It still happens with even the most advanced AGIs (Artificial General Intelligence) tools, such as Open AI's ChatGPT and Google's Bard. We’re developing ways to design around this issue and are sharing a showcase soon.

Another key factor is to continuously track the questions and answers the AI is giving to learn from its real-life usage. The idea is to continuously improve the AI by feeding it with enhanced knowledge to respond with even more valuable answers in the future.

In a nutshell

AI brings exciting new opportunities to the web. As designers and developers, we should use it responsibly to help people navigate — and craft the AI’s knowledge, personality, and design to create fantastic experiences.