Humans, humans everywhere. Notes from Converge 2026

At Converge 2026, design system speakers proposed firm automatic checks for AI agents and named human owners for decisions, according to attendee Cassie Groos.

Cassie Groos, a design system practitioner, summarised the nine talks at Converge 2026, a design system conference run by the documentation company Zeroheight in Newcastle. Speakers described how AI tools have raised the volume of generated interface work and the drift away from shared components. Their common answer was clear automatic rules for AI agents, better documentation for agents to read, and people who own the decisions.

• Dan Donald, one of the speakers, sorted guidance into three kinds. Rules are checked, guidelines are reviewed by people, and decisions are owned by a named person.
• Zeroheight's founder, Jerome de Lafargue, described an AI agent that notices Figma changes and proposes documentation updates, which a person approves before they merge.
• Frances Szweda, speaking from her work at the retailer Tesco, argued for design system advocates at every level of the organisation. One of her slides said that a user's behaviour takes 8 to 12 weeks to change.
• Hannah Clarke, another speaker, warned that AI tools can reduce design system use, because people can generate their own components. She proposed documentation served to AI agents plus scripted safeguards.

A design system team can decide in advance which guidance AI agents must follow, which a person reviews, and which decisions belong to a named owner. This is a second-hand summary of conference talks by one attendee, not a report from a single team's practice.

Derived by Working Surface from the article; more in the Playbook. Source line: However much the agents do, somebody still has to decide.

26 October 2026: teams now judge AI work by tests a person sets, from pass-or-fail checks on what agents build to plain questions about who asked for a feature.