Designers direct the system.
AI compresses the groundwork.

• Cross-domain auditing — systematic review of multiple TA domains and sub-properties to surface recurring UI patterns, shared foundations, divergent behaviours, and inconsistencies the system will need to reconcile.
• Foundation extraction & organisation — typography (scale, families, weights, line-height, responsive behaviour), colour systems (brand, semantic, surface, state, accessibility-paired), spacing, layout grids and breakpoints, elevation and shadows, radius, responsive and adaptive patterns, motion and interaction primitives, iconography, localisation (multi-script, RTL readiness, content expansion), and accessibility baselines.
• Component intelligence — identification of high-frequency components doing the majority of UI heavy lifting across domains, including variant patterns, states, and contextual usage. Informs prioritisation of what gets systematised first.
• Token & variable architecture — assistance setting up a scalable Figma variable and token architecture: primitive → semantic → component-level layering, with mode/theme structures anticipating future expansion (dark mode, sub-brand theming, campaign overlays).
• Taxonomy & documentation scaffolding — support for naming conventions, taxonomy structures, and documentation frameworks so the system speaks a consistent language across Figma, code, and written guidance.
• Where the time savings go — designer effort is redirected toward governance, accessibility, UX quality, system thinking, pattern coherence, and long-term maintainability.

• Manual validation of extracted foundations, tokens, and component inventories against source domains.
• Design + Dev review for token accuracy, naming, and implementation feasibility before adoption.
• Governance checkpoints before any pattern enters the live system.
• Version-controlled variable architecture, naming conventions, and documentation; changes require an ADR.
• Accessibility verification — contrast, focus, semantic structure, WCAG conformance — on every foundation and component before sign-off.
• Source traceability — every AI-surfaced pattern links back to the domains and instances it was derived from; uncited patterns flagged for rejection.
• Bias & over-fitting check — designers verify AI hasn't over-weighted recent redesigns, the most visually prominent property, or surface-level repetition mistaken for system intent.
• Human approval gate — no foundation, token, or component goes live without designer sign-off.

• Manual cross-domain audits
• Manual token and foundation extraction
• Manual component inventory creation
• Manual Figma variable and library setup
• Manual documentation authoring and governance tracking

Sources · TA Client Dashboard — timeline & key dates ↗ · CMS Migration — sprint plan & deliverables ↗

• Scope pollution — a large share of variables are still on ALL_SCOPES, which clutters every property picker. They should be narrowed to their real use (fills, text, gap, radius…).
• Empty component-token layer — the 08 Component collection exists but holds no variables; component tokens haven't been defined yet.
• Sparse component documentation — only — of — component families carry a written description; naming and usage guidance need to catch up to the build.
• Sound layering — tokens are aliased rather than held as raw values, and theming is cleanly separated across brand modes, locales and breakpoints. The architecture is holding up.