Dark Mode

The Problem

Visible request: 1,200+ users requesting dark mode for accessibility and eye strain relief

Hidden reality:

• Outdated theming architecture with thousands of lines of inconsistent styling
• Low design system adoption across 22+ teams building in silos
• 224 accessibility bugs system-wide
• No semantic color system - hard-coded values everywhere
• Every new theme would compound technical debt

The question: Ship dark mode fast, or use it as the vehicle to fix years of foundational problems?

The Decision: Fast Path vs. Trojan Horse

Led a strategic offsite to evaluate approaches:

Option 1: VP Product proposal - Hard-code dark mode values

• Ship what users want, ship it fast
• No coordination complexity across 22 teams
• Increases technical debt
• Bug-prone (every hard-coded color is a maintenance burden)
• Doesn't leverage design system
• Every future theme compounds the problem

VP skepticism: "Can you really coordinate 22 teams for this? Just ship it."

Option 2: My proposal - Dark mode as trojan horse ✅

• Introduce new theming architecture as a breaking change
• Force all teams to migrate to modern design system patterns
• Teams can't access dark mode without using proper semantic tokens
• Fix foundational technical debt WHILE shipping user-facing feature
• Use dark mode as the compelling carrot that drives modernization

The bet: Dark mode would be compelling enough that teams would prioritize migration work.

The forcing function: no proper tokens = no dark mode access. And a11y.

How I Made It Happen

Resourcing strategy:

• Attached dark mode to a higher-priority initiative to get it resourced
• Established tiger team for system-wide polyfill requirements

Execution:

• Coordinated 22 teams across multiple geographies
• Created migration playbook for all 22 engineering teams
• One quarter timeline for complete migration
• Introduced semantic color variable system
• Upgraded entire design system to v9
• Made dark mode the unlock for design system adoption

Key approach: Made the new system SO much better that teams wanted to migrate, not just had to migrate.

The Impact

User adoption - Fastest in Zendesk history:

• 15K users in first 2 weeks
• 60K+ users in under 3 months
• November `25 100K+ agents using dark mode daily

Platform improvements:

• 40% increase in design system adoption across all teams
• All comboboxes and dropdowns replaced with modern components
• Entire design system updated to v9 with hundreds of fixes
• All colors migrated to semantic variable system
• Fixed 224 accessibility bugs (43% of all system-wide issues)

Business outcome:

• Delivered on top #10 user-requested feature
• Modernized frontend architecture for faster future development
• Set pattern for how platform improvements get prioritized

CPO's reaction: First increased scope 1 month before launch, and when we did it he laughed and said, "Okay, you win."

Key Learning: Using User Demand as Platform Leverage

The principle: Platform work needs a user-facing hook to get prioritized in reactive organizations.

What didn't work:

• Writing vision docs about "we should modernize our theming"
• Making the case on technical merit alone
• Asking teams to migrate without compelling reason

What worked:

• Attaching platform work to something users were demanding
• Making the improvement a forcing function (no migration = no feature access)
• Creating compelling enough value that teams wanted to do the work
• Framing it as competitive advantage, not just technical improvement

The pattern: Find the user-facing feature that requires the foundational work you need to do. Use that feature as the vehicle to drive architectural improvements.

Team & Execution

Cross-functional coordination:

• 22+ engineering teams across multiple geographies
• Design system team
• Accessibility team
• QA and testing
• Product teams across Support, Admin, Guide, and acquired products

Timeline: One quarter for complete system-wide migration

Key to success: Clear migration path, comprehensive documentation, tiger team support for complex cases