Outcome
A shared definition of design readiness and design done — giving designers, Product and partners a clearer way to expose missing decisions, unresolved UX risk and quality gaps before they became expensive to fix.
Overview
Under delivery pressure, “done” quietly shrinks: accessibility gets deferred, edge cases become follow-up tickets, empty and error states arrive late, and design enters after important product decisions are already made. I built Design Gates to make those gaps visible earlier — two checkpoints with a written UX Quality Bar between them: Entry (is the problem defined well enough for design to begin?) and Exit (is the experience complete enough to hand over responsibly?). The bar gave designers, contractors, Product and Engineering a shared definition of done, used before Figma opened, during reviews and before delivery.
Why this was needed
With 50+ initiatives moving in parallel and a stretched design team, one recurring failure mode was late design involvement: solutions arriving with scope already defined and too little time left to resolve fundamental UX problems. The risk was rework, missing states, accessibility gaps, unclear acceptance criteria and unresolved product decisions travelling into build. The operating model moved those conversations earlier.
- Readiness Enough evidence, scope and clarity to design responsibly.
- Ownership One named design owner per initiative.
- Quality An explicit definition of done, not individual judgement alone.
- Risk Unresolved gaps surfaced while teams can still act.
The system
Gate 1 · Ready for design
Design should not begin until there is enough definition to design against. The entry check asks whether the PRD or working-backwards intake includes: problem and user; outcome and baseline; scope and exclusions; flows across happy, empty, error and edge conditions; technical, regulatory or operational dependencies; acceptance criteria; and what is known versus still assumed. If critical information is missing, the work goes back for clarification rather than entering Figma with ambiguity disguised as progress.
UX Quality Bar
The full UX Quality Bar has ten sections. Seven of its checks set the minimum release bar for design done:
-
01
Accessibility
WCAG AA considered and validated.
-
02
Design system
Global Design System applied; deviations intentional and reviewed.
-
03
Edge cases
Scenarios beyond the happy path accounted for.
-
04
Empty states
Clear next action where one is needed.
-
05
Error states
Failures explained with a recovery path.
-
06
Copy
Reviewed for clarity, consistency and context.
-
07
Success
Criteria explicit and measurable.
These sit inside a broader design-review lens covering problem clarity, flow, information architecture, cognitive load, responsive behaviour and interaction quality. The seven criteria define the release bar; the broader lenses ask whether the solution is actually good.
Gate 2 · Design review
The exit review uses three outcomes: Pass — bar met, ready for handoff; Risk — gaps remain, proceeding only with risk named, owned and time-bound; Fail — bar not met, work returns to design. The language makes ambiguity harder to hide: a review should not end with “looks fine” when known product risk still exists.
Why PASS / RISK / FAIL worked under delivery pressure
When teams lack a shared definition of done, review becomes subjective. PASS / RISK / FAIL shifts the question from “Do we like this?” to “What remains unresolved, how serious is it, and who owns it?” — so risk is accepted consciously rather than disappearing into delivery. With dozens of initiatives at once, the same readiness checks and review language had to work for in-house designers, contractors, Product and Engineering. By the end of 2025 it was a more unified way of working — less quality dependent on memory, seniority or individual heroics.
How AI fits
AI-assisted checks feed the same system rather than a separate definition of quality. Automation can flag missing states, design-system inconsistencies, accessibility concerns, content issues and review prompts — but the output still enters PASS · RISK · FAIL. AI can make inspection faster; it does not own readiness.
Outcomes
- Shared entry standard — problem, scope, evidence and acceptance criteria before design begins.
- Visible ownership — one design DRI through definition, review and handoff.
- Earlier UX involvement — design in before product and funding decisions lock.
- Written quality bar — accessibility, system use, states, copy and success made explicit.
- PASS · RISK · FAIL — shared language for unresolved UX risk.
- More consistent delivery — same standard across permanent and external capacity.
- AI-ready governance — automated review feeds the existing quality system.
Reflection
The most useful quality systems are not the most complicated ones — they make important questions difficult to ignore. The gates do not replace judgement, collaboration or strong craft; they give those things a clearer place to happen.