AI agents are now users. Nielsen Norman Group argues that the foundational design assumption, that 'user' means 'human', is no longer accurate. Agents navigate websites, fill forms, compare options, and execute transactions on the same interfaces built for people.
They do it badly. NN Group is direct: agents operate crudely, unreliably, and with significant limitations. That caveat matters because it reframes the design problem. You are not optimizing for a capable autonomous system. You are designing for something that fails in ways human users do not, which forces a rethink of interaction patterns, error states, and interface structure.
The full article goes deeper into what it actually means to design for a non-human actor sharing space with human ones, and why accessibility principles are the practical starting point. If you design any transactional or navigable interface, this is the conceptual update your mental model needs now, not later.
[READ ORIGINAL →]