September 29, 2026

Updating the Heuristic Evaluation

My journey into designing and developing my own product

UX DesignProduct Design
Updating the Heuristic Evaluation

When designing my product, I wanted to implement as many heuristics for my users as possible. I wanted the user to find real value in my product, and I did not want it to exist just to make money from my users. While designing, I kept thinking about all the heuristics of design. I remember working on past projects, doing heuristic evaluations and writing reports. Now we are faced with a new kind of product: the AI-backed product. The AI-backed product works a bit differently from the traditional apps and software we are used to. It has different mechanics and different rules, so our heuristics need to work for this new generation of products. The question it comes down to is how to balance what is good for the user with the interests of the business. How do we empower the user without causing chaos?

Jakob Nielsen's third usability heuristic is User Control and Freedom. People make mistakes, and when they do, they need a clear way out, an emergency exit, and the freedom to undo and redo. It is what keeps the user in control of their product. It is also a challenging heuristic in the face of AI. When a product is built to be fully automated and controlled by a corporation, how can I, a small product designer, give my users freedom and control over their product while avoiding crashing the system and causing chaos? I am not the first to ask this. In 2019, researchers at Microsoft published a set of guidelines for designing with AI, and one of them is to let users customize what the AI pays attention to and how it behaves. The question is how to do that well.

Giving control to users needs to be done responsibly. In a vehicle, we give the driver a steering wheel, brakes, and an accelerator to control their vehicle. So how do we achieve control and freedom in AI products? In my product, I gave my users control over the soul of the product, in a way they can understand and in a way that will not cause chaos. In Friendly, the user can shape the soul of the product by shaping the memory of their AI companion. The user has the freedom and control over what their AI companion knows about them, and can shape it the way they like. If their companion remembers something they wish it did not, they can change it or remove it. That is my emergency exit, and my undo, for the age of AI.