01
Sound familiar?
"We have three products and one inconsistent UI library."
"Our AI prototypes never match our design system."
"The same button exists four times, each slightly different."
"Handoff still means screenshots and guessing."
Book a call
ENSK
A system your team and their AI build from.
Agents assemble UI from the same components as your team. When the system is not machine-readable, they guess. We build systems where they do not have to.
02
The layers.
01
Token architecture
Three layers: primitive, semantic, component. A closed vocabulary in the DTCG standard format. Agents pick values instead of inventing them.
02
Components
Built with props, variants, states and use cases. Documented for people, structured for agents.
03
Component metadata
Props, variants, use cases and anti-patterns as structured data, not prose.
04
Design patterns
How components compose into flows and screens, so the system answers questions before they are asked.
05
Motion
Durations, easings and transition patterns defined in the system, not invented per screen.
06
Agent distribution
Always-on rules for foundations plus on-demand lookup, delivered as structured files your agents read.
07
Machine-readable governance
Versioning, deprecations and changelogs agents can read, so the system stays true over time.
We design the system and its documentation. Your engineers implement it in code. The handoff is built exactly for that.
Agents building from pre-built components reached 95% design system compliance. With style-guide instructions alone, 71% (Cha et al., CHI 2026).
This is design for AI-assisted teams. AI consumes the system; people design it.
03
The ladder.
01
Full audit
02
Retrofit an existing system
03
Build from scratch
Scope confirmed on a call.
04
What you get.
A token set in the DTCG format, three layers deep
Components with props, variants and states
Structured metadata, not prose docs
Design patterns for the flows that matter
Motion: durations, easings and transitions as part of the system
Governance documentation your team can run
Adoption plan and handoff for your engineers
05
How it works.
01
Full audit
Where your product and its system stand against the layers.
02
Foundations
Tokens, naming, structure. The closed vocabulary.
03
Components and metadata
Documented for people, structured for agents.
04
Distribution and adoption
Rules, documentation, handoff. The system becomes the default.
06
System governance.
A system stays useful only while someone keeps it true to the product. After launch, we run it with your team.
01
Maintenance
Tokens, components and documentation kept in line with the product.
02
Review
We check the UI your team and their agents produce against the system and log what drifts.
03
Updates
When the product or your tools change, the system and its agent files change with them.
04
Monthly release
One release a month, with a changelog people and agents can read.
Iterate every month, from €800/month.
Questions, answered.
What does agentic-ready mean?
Structured so AI agents can read and use it: closed vocabularies, structured metadata, documentation agents can parse. People design the system. Agents consume it.
Does this replace our designers?
No. It makes their decisions the default everywhere, including in agent output.
We already have a design system.
Most clients do. The ladder starts with a full audit; retrofit is the common path.
How long until agents can use it?
Depends where you start. A full audit answers that with a plan, not an estimate.
Do you deliver code?
No. We deliver the system: tokens in the DTCG format, ready for code, and component specifications with props, variants, states and use cases. Your engineers implement the components.
How much does an engagement cost?
We work as a fixed project or an ongoing partnership, scoped and priced before we start. Engagements typically begin with the audit, from €2,500. Continue into design within 60 days and the audit is included in the design price.