Systems that remember
Why modern design systems aren’t libraries, they’re organisational memory.
TL;DR — Key Takeaways for Time-Poor Execs
- Modern design systems succeed by reducing the cost of change, not by adding more components.
- 2025 systems rely on governance, observability, and AI to detect drift before users ever feel it.
- Metrics that matter now tie directly to risk, compliance, velocity, and quality.
- The future isn’t “more components,” it’s adaptive organisational intelligence that remembers, predicts, and governs.
- Teams win when their systems retain decisions, surface drift, and make change safe.
1. The Post-Hype Era
Modern systems prioritise memory, resilience, and cost-of-change.
Every few months, another manifesto claims to reinvent design systems. But the real issue in most organisations isn’t tooling, it’s the rising cost of change.
Imagine a billion-dollar startup attempting a full-system migration over a weekend. The fallout is predictable: duplicated components, broken flows, emergency rollbacks, and a week of UX outages. The failure isn’t technical, it’s the absence of organisational memory.
The most effective systems today aren’t faster because they’re bigger.
They’re faster because they remember what worked, observe where drift begins, and lower the cost of change across the organisation.
From Parts → Pipelines → Permanence.
This diagram illustrates how modern design systems move beyond isolated components toward persistent, observable infrastructure that reduces the cost of change.
2. From UI Kits to Ecosystems
How design systems expanded from files to organisational platforms.
Design systems matured through distinct eras, not by accident, but by accumulating capability.
The evolution:
- 2018–2020 — Standardisation
Shared components replaced chaotic files. - 2021–2023 — Tokenisation & Platformisation
Tokens unified design → code → CI/CD → analytics → compliance. - 2024–2025 — System-as-Product
Observability, governance, and automation turned systems into operational infrastructure.
Accessibility/performance compliance (EU A11y Act, state digital standards) shifted quality from “nice to have” to risk mitigation.
Observability turns assumed value into proven value.
The transformation of design systems over the last decade, showing how each era added new layers of capability, from shared components to fully governed, intelligent platforms.
3. What’s Actually New in 2025
The shifts redefining system architecture and workflow.
2025 brings two kinds of changes: core shifts everyone must adopt, and edge shifts early movers are exploring.
Core shifts
- Semantic token hierarchies
- Figma Variables as the backbone
- W3C-aligned design tokens
- AI-assisted authoring
- Governance baked into CI/CD
Edge shifts
- Composable, multi-brand systems
- Partner ecosystem compatibility
- System observability dashboards
- Connected documentation
- Accessibility & behavioural tokens
Core shifts = mandatory.
Edge shifts = competitive advantage.
Modern design systems function as continuous feedback loops, where telemetry and documentation influence the token layer that powers every component and app.
4. Myths That Need to Die
Persistent misconceptions that slow organisations down.
| Myth | Reality |
|---|---|
| “You only need 20 tokens.” | You need 20 core tokens—but hundreds of semantic layers at scale. |
| “Headless = better.” | Headless without guardrails creates exponential inconsistency. |
| “Tokens = scalability.” | Scalability requires governance + observability, not just tokens. |
| “Maturity = conformity.” | Maturity = controlled divergence within guardrails. |
| “AI replaces designers.” | AI is the junior designer. Humans still make the decisions. |
Five pervasive myths that still mislead organisations, contrasted with the realities required for scale, governance, and long-term consistency.
5. The Pillars of a Modern System
Eight foundations that make change safe, observable, and scalable.
1. Governance & versioning
- KPI: % of components with a defined owner
- Example: Team G reduced approval cycles from 9 days to 3 by assigning clear ownership.
- KPI: Mean Time to Update (MTTU) after a spec change
- Example: Team R cut MTTU from 14 days to 48 hours after adding version gating.
2. Accessibility by design
- KPI: % of components passing automated a11y tests
- Example: Team M pushed coverage from 62% to 94% after adding a11y tokens.
- KPI: % of flows supporting reduced-motion
- Example: Team B eliminated 100+ unsafe animations in two sprints with motion tokens.
3. Testing & quality assurance
- KPI: % of components covered by visual diff testing
- Example: Team L dropped regressions 43% after hitting 80% diff coverage.
- KPI: Regression rate per release
- Example: Team Q halved regression incidents by adopting AI diff reviewers.
4. Cross-platform tokens
- KPI: # of platforms powered by the same token source
- Example: Team E unified web/iOS/Android styling with one token pipeline.
- KPI: # of manual overrides required
- Example: Team F went from 147 overrides to 12 after refactoring semantic tokens.
5. Adoption & education
- KPI: Reuse ratio (% UI built with system components)
- Example: Team H boosted reuse from 38% to 71% through better onboarding.
- KPI: # of system contributors per quarter
- Example: Team D doubled contributors after running monthly “Component Clinics.”
6. Observability
- KPI: Time to detect design drift
- Example: Team A cut drift detection from 10 days to 2 by dashboarding usage.
- KPI: Component health score trend
- Example: Team K improved failing components from 22% to 3% using health scoring.
7. Sustainability
- KPI: % of key flows meeting CLS/LCP targets
- Example: Team J met CLS targets across 95% of surfaces after adopting budgets.
- KPI: % of components adhering to motion/performance budgets
- Example: Team V reduced CPU-heavy animations by 80%.
8. Composability
- KPI: % of UI built from shared primitives
- Example: Team O raised primitive usage from 40% to 89% after rationalisation.
- KPI: Net change in total component variants
- Example: Team C eliminated 63 redundant variants in one quarter.
The structural pillars that make change safe and predictable, supported by practices that ensure systems remain healthy, inclusive, sustainable, and widely adopted.
6. Metrics That Actually Matter
System value is proved through velocity, quality, and resilience.
- MTTU
- Example: One team cut MTTU from 3 weeks to under 2 days with automated releases.
- Reuse ratio
- Example: A product unit reduced dev time 30% after raising reuse by 20 points.
- Design-to-dev latency
- Example: A fintech team halved latency by linting PRs against tokens.
- Component health score
- Example: A retail org moved 70% of components from “red” to “green” in one quarter.
- Design-debt velocity
- Example: A platform team resolved 90% of drift tickets within the same sprint.
- Regression reduction
- Example: A system with AI diffing saw 45% fewer regressions.
7. AI’s Role — From Generator to Governor
AI accelerates; governance keeps it safe.
Generator
AI scaffolds variants, tokens, and component structure.
Critic
AI audits naming, accessibility, and architectural consistency.
Governor
AI blocks PRs that violate tokens, a11y rules, or performance budgets.
Gov/Compliance Note
Models must follow explicit guardrails and rely on approved internal patterns, not public scrapes, for IP safety and traceability.
AI now contributes across the design system lifecycle, but governance keeps it safe, ensuring AI-generated changes respect tokens, accessibility, and performance constraints.
8. The Design System Maturity Framework
A ladder from UI kit → intelligence.
| Level | Focus | Example KPIs + Anonymized Examples |
|---|---|---|
| 1 — UI kit | Shared library | ≥70% reuse • Team Z hit 72% reuse after standardising primitives. |
| 2 — System | Tokens + docs | Docs linked to code • Team B synced docs + code and cut onboarding time by 40%. |
| 3 — Product | Pipelines | MTTU <72h • Team R shipped updates twice weekly after adding CI/CD gates. |
| 4 — Platform | Multi-brand | ~90% reuse • Team Y unified three brands with one semantic layer. |
| 5 — Culture | Rituals | Cross-team contributors • Team P tripled contributors via monthly sessions. |
| 6 — Intelligence | Predictive governance | AI flags drift • Team A caught mismatches pre-merge instead of post-release. |
Design systems mature through six stages, from basic component libraries to predictive, intelligence-driven platforms that anticipate drift before it happens.
9. The Human System Behind the System
People—not tooling—determine longevity.
- Triads create shared ownership.
- Narratives position the system as a win, not a restriction.
- Rituals like Migration Days prevent silent drift.
- Cross-org collaboration ensures brands can extend safely.
Systems stabilise work.
People sustain systems.
10. Tokens for Reality: Localisation, Motion, Ethics, Context
Systems must adapt to language, body, and environment.
- i18n tokens for spacing, line height, mirroring
- Motion & sound tokens for comfort and accessibility
- Context-adaptive tokens for density, platform, input types
- Ethical tokens for cognitive load and animation limits
Token layers evolve from simple variables into adaptive structures capable of handling language, motion, context, and ethical constraints across platforms.
11. Sustainability & Performance as First-Class Citizens
Performance is now tied to compliance, risk, and reputation.
Regulations increasingly make performance and accessibility mandatory, not aspirational.
Design systems now shoulder:
- Performance budgets
- Example: Team C met CLS/LCP budgets across 92% of pages after implementing them at the system layer.
- Energy-aware defaults
- Optimised delivery via tree-shaking & usage-based code splitting
Better performance = less risk, less churn, and more trust.
Performance and energy usage are now first-class observability metrics, helping teams meet compliance requirements and reduce system-level risk.
12. Where It’s All Headed
Design systems are evolving into organisational intelligence.
Systems are becoming knowledge graphs:
they understand how tokens → components → flows → business rules connect.
Takeaway:
Design systems are shifting from static assets to adaptive organisational memory.
13. Futurewatch
Signals shaping the next generation.
- AI governance agents
- Behavioural & ethical tokens
- Cross-enterprise composability
Takeaway:
The next generation of systems won’t just respond, they’ll anticipate.
14. After the System
Why systems that remember build teams that win.
A system that remembers reduces cost.
A system that observes prevents drift.
A system that adapts earns trust.
If your organisation’s design memory is strengthening—or slipping—I’d love to hear what you’re seeing.
10 Free Repeating Chrome Tribal Patterns
Free for non-commercial use, but don’t even think about using them for your corporate schemes.
Designing the Pause Between Clicks: 11 Patterns That Help Users (and Teams) Exhale
After years designing interfaces for finance, health, and global brands, one thing’s clear: Good d
Measure What Moves Users
A playbook for UX metrics that actually change the business Executive Summary If you’re tired of m