Terminology in the field
The vocabulary is fragmented because sources describe different objects: designing AI experiences, enabling agents to consume a design system, or designing software that can act. Terms should not be treated as synonyms.
| Term | Where / who uses it | Recency | Apparent meaning | Status | Related terms and evidence |
|---|---|---|---|---|---|
| Carbon for AI | Formal IBM Carbon name | active 2025–26 | An extension of Carbon for marking AI presence, explainability, AI variants, chat, tokens, code and Figma assets | Established organizational language; unusually concrete | Carbon for AI |
| AI label | IBM Carbon; formerly “AI slug” | stable 2026 | Interactive indicator and pathway to explainability | Established component name within Carbon, not industry-wide | AI label |
| AI and Joule Design | SAP Fiori | active 2025–26 | Guidance and patterns for generative/intelligent SAP experiences | Formal organizational language | SAP guidelines |
| Rovo Dialect / Rovo UI | Atlassian Design | active 2026 | The AI “voice” nested within Atlassian’s design language: principles, foundations, components and patterns | Formal, brand-specific language | Rovo Dialect |
| Copilot UI Kits | Microsoft Fluent | active 2026 | AI-focused components and patterns extending Fluent across web, iOS and Android | Formal organizational language | Fluent design resources |
| content engineering | Microsoft Fluent | active 2026 | UX/content practice for shaping model behavior using roles, tasks, rules and examples | Formal practice term, not a synonym for AI-ready DS | Fluent |
| AI-first design system | Microsoft Design | 2026 | Direction toward a coherent Copilot interaction architecture | Emerging organizational framing | A simplified system |
| AI-native design system | Atlassian; practitioner and vendor discourse | strongly recurring in 2026 | At Atlassian: DS as semantic context/infra for humans and AI, with AI patterns included. Elsewhere often vague marketing | Emerging; meaning unstable | Atlassian, community example |
| context engine / context layer | Atlassian | 2026 | Structured content, MCP, skills, templates and semantics that let agents understand and build with ADS | New but backed by formal organizational work | Atlassian |
| Compositional Design System | SAP | 2026 | Explicit relationships, semantics, constraints and guardrails for dynamically composing coherent interfaces | New formal term; production claim in Joule Work, but organization-specific | SAP |
| generative UI | SAP, research, tool and practitioner discourse | accelerating 2025–26 | Interfaces assembled or adapted dynamically from intent/context, rather than fixed screens | Recurring research/product term; implementations vary greatly | SAP, MCP Apps |
| design-system-aware AI / design-system-aware development | Research/practitioner wording | 2026 | AI generation grounded in a named DS rather than generic components/styles | Emerging research term; useful and comparatively precise | 2026 evaluation |
| machine-readable design system / machine-readable context | SAP, WordPress, practitioner discourse | 2026 | Components, tokens, examples, semantics and rules exposed in structured forms agents can query | Recurring descriptive language, not a standard | WordPress MCP, SAP |
| design system MCP | Carbon, WordPress, Storybook, Figma, zeroheight and practitioners | dominant new implementation phrase in 2025–26 | An MCP interface that exposes DS documentation, manifests, components, tokens, examples or tools to agents | Rapidly recurring implementation term; protocol is established, DS payload conventions are not | Carbon MCP, Storybook MCP, Figma |
| agent-ready design system | Into Design Systems, practitioner blogs/forums | 2026 | A DS whose knowledge and implementations are accessible and actionable to coding/design agents | Emerging practitioner language; no accepted checklist | Into Design Systems, De Voorhoede |
| agentic design system | Conference/practitioner/vendor language | 2026 | Usually machine-readable DS knowledge plus agent workflows; sometimes implies the DS itself performs work | Emerging and ambiguous; often marketing | TDP, Into Design Systems |
| agent-native design system | Small commercial libraries | 2026 | Component registry designed to be invoked by coding agents | Isolated/marketing-heavy | AI2 |
| agentic UX / agent design | Microsoft, research, Smashing/ServiceNow practitioner writing | strongly recurring 2025–26 | Interaction design for systems that plan, use tools, and take actions with variable autonomy | Established umbrella in current discussion; detailed patterns are emerging | Microsoft, Smashing |
| human-centered design for agents | Microsoft | 2026 | Lifecycle guidance from first-run through interaction, correction and recovery | Formal organizational language | Microsoft Learn |
| AI-ready design system | zeroheight marketing/survey, independent guides | 2025–26 | Usually “usable by AI tools,” sometimes broadly “supports AI products” | Recurring but not defined; frequently marketing shorthand | zeroheight report, DesignSystems.one |
| AI-aware / AI-optimized design system | Scattered practitioner/marketing usage | 2025–26 | Generally optimized documentation/workflows or awareness of AI-generated states | Isolated; no stable meaning | No strong primary organizational definition found |
| AI design language | Broad product/brand usage | ongoing | Visual and behavioral language for AI experiences | Meaningful within a named organization, too broad across industry | Compare Rovo Dialect, Carbon for AI, Fluent Copilot |
| AI interaction patterns / AI UX guidelines | SAP, Google PAIR, Microsoft HAX, Atlassian | established and current | Guidance for uncertainty, feedback, explainability, control and errors | Most established cross-organizational vocabulary | SAP, PAIR, HAX |
| AI components / AI component library | Carbon, Fluent, SGDS, vendors | 2025–26 | Shipped components for AI disclosure, chat, generated content or branded AI moments | Concrete when tied to code; otherwise marketing | Carbon, SGDS |
Terms discovered beyond the starting list
- Context engine / context layer is the clearest new organizational phrase for the “AI builds with the DS” problem.
- Compositional Design System, composition logic, and experience models distinguish generative UI from ordinary component retrieval.
- Rovo Dialect shows one organization treating AI expression as a nested design-language dialect, not a universal AI layer.
- Content engineering moves part of AI quality into the design system’s content/behavior discipline.
- AI presence is a concrete state or modifier in Carbon, more precise than “AI component.”
- Shared system context (SGDS/Figma) describes continuity between design, code and agents without claiming that the DS itself is native to AI.
Assessment
There is no common industry term for the whole space. “AI-ready” and “AI-native” recur, but collapse distinct goals and are frequently promotional. The most stable vocabulary is problem-specific: AI UX guidelines/patterns, named organizational extensions such as Carbon for AI, design system MCP for agent access, machine-readable context for structured knowledge, generative UI/composition logic for dynamic assembly, and agentic UX/agent design for action-taking software.