Don't make me prompt

The blank prompt box is a terrible host. It sits there, blinking, and waits for you to find the right words.

In my first article, Don’t Make Me Prompt, I collected eight interaction patterns that replace the prompt with something better: suggestions, direct manipulation, voice, and AI that works inline. Since then I have used these patterns in talks, workshops and courses with designers, developers and product managers. One observation keeps coming back. Teams fix the surface and leave the structure untouched. They add suggestion chips or a microphone button, and the product still fails its users.

The prompt box is only the visible part of the problem. Below it sits a set of design decisions: what the AI should do at all, how it explains itself, who stays in control, and what happens when it is wrong.

I teach these decisions as the 6 HAX Principles (Human-AI Experience). I did not invent them. I curated them from the leading work in the field: Microsoft’s HAX Toolkit and the research behind it (Amershi et al., 2019), Google’s People + AI Guidebook, Apple’s Human Interface Guidelines for machine learning, the pattern library here at aiverse, and the research literature on human-AI interaction. Then I validated them in client work in pharma, healthcare, fintech and logistics. Today they are the core of what I teach at hax.academy.

This article walks through the six principles. For each one I point to recordings in the aiverse library, so you can watch the principle at work in a real product.

Why the prompt is still the problem

Start with the numbers. In a study with 37,000 volunteers, people typed 36 words per minute on mobile devices (Palin et al., 2019). In a controlled comparison on touchscreen phones, speech input was almost three times faster than the keyboard for English: 153 against 52 words per minute (Ruan et al., 2016). A tap on a button needs no words at all.