NN/G: The four design jobs AI created (so far)
Most discussions of AI and design jobs frame the question as replacement — how many roles will disappear. This June 2026 article from Nielsen Norman Group researcher Sarah Gibbons takes a different angle. It maps four new categories of design work that AI has created and asks where demand is outpacing supply.
The article is structured around four orientations, each requiring distinct expertise.
The four orientations
Designing with AI — using AI as a tool within an existing design workflow to accelerate ideation, prototyping, copy, and research. This is where most designers currently operate, and it builds naturally on existing skills.
Designing AI products — building the interfaces and experiences that deliver AI capabilities to users. This splits into two groups: AI-native products built entirely around AI, and AI features embedded in existing products. Interface patterns here are still evolving, and there is no established playbook.
Designing for AI agents — structuring the content, data, and interaction flows that autonomous systems parse and act on. Gibbons calls this “designing the roads” that agents travel: data organization, interface scaffolding, and information architecture optimized for machine rather than human navigation.
Designing the AI — shaping model behavior, evaluation criteria, and system principles. This includes defining what the AI refuses, where it expresses uncertainty, how it weighs conflicting instructions, and what constitutes a good response.
Where the gap is
The article makes the gap explicit: demand for the last two orientations — designing for agents and designing the AI — is growing faster than the supply of designers who can do them. Most organizations still leave these decisions to engineers by default, not because engineers are better placed to make them, but because no designer with the right expertise is available.
Who this is useful for
The article is useful for designers thinking about where to specialize next, for design leaders auditing whether their team covers all four orientations, and for hiring managers writing job descriptions that go beyond generic “AI designer” language. It reads quickly and makes concrete distinctions that are easy to act on. What it does not do is tell you how to develop expertise in any particular orientation — it names the territory without providing a map.