Your design process now has a feedback loop built in

AI agents that pause, check their own output and retry are starting to change how iterative work gets done — including design critique loops. It's a shift in method, not just tooling.

A widely-shared 2026 piece on agentic workflows makes a simple point: agents that reason, adapt and self-correct are different from old-school automation that just runs a fixed script. Applied to design, that's the difference between a plugin that resizes frames and one that checks its own output against your design system before handing it back. For methods, the interesting bit isn't the tech — it's what it nudges you to do differently. If an agent can draft and self-critique a first pass, your value shifts further toward framing the right problem, setting constraints, and making judgement calls an agent can't (should we even build this?). Labelled adoption: agentic patterns are showing up across finance, legal and ops workflows already, even if the design-specific version is still forming.

Related in UCD methods

Adoption

Turning workflow diagrams into testable agent flows

Vertesia's July 2026 product update shows an AI agent converting a photo of a workflow diagram into a testable, interactive process, then running every path through it before anything ships. It's a live example of AI compressing the gap between planning and working prototypes.

vertesiahq.com22 Jul

Hype

2026 trend: designers push back against AI-slick

A 2026 design trends roundup points to imperfection, distortion and human craft as a reaction against visual sameness from AI-generated design. It's framed as commentary and early aesthetic signal rather than settled practice.

lindsaymarsh.substack.com15 May

Established

Usability testing 101 still holds up in the AI era

Nielsen Norman Group's core usability testing guidance is decades old — and still the fastest way to check if an AI-generated or hand-crafted design actually works. Foundational, not fashionable.

nngroup.com16 Jan