Declare what needs review.
Choose one primary category, an optional secondary category, and make known placeholders visible.
- Visual & Interaction Fidelity
- Content & Language
- Connectivity & Flow
- Informational Structure
Design leadership / AI Governance
AI, with accountability.
Asked to share lessons from Rock This, my independent project, I developed a proposal for how Citi Design could adopt AI with clear standards, shared ownership, and human judgment.
01 / Review
From prototype to product.
Partners were submitting polished AI prototypes through intake. The work still needed the knowledge, constraints, and decisions of other disciplines.
Example handoff
Primary: Connectivity & Flow
Secondary: Visual & Interaction Fidelity
Intent: Validate checkout step order and branching.
Known placeholders: Draft copy; simulated data.
Primary feedback is blocking for this review; secondary feedback is advisory.
Bring in the right disciplines. Record gaps and route unresolved questions to Product, Design, Research, Content, Accessibility, Engineering, and Risk as needed. A focused review does not establish product approval.
02 / Access
Access grows with evidence.
Capability, project need, and partner review determine the environment and permissions.
Tier 0 / AI-literate
Approved tools. No sensitive data.
Tier 1 / Promptable systems
Design-system assets. Owner review.
Tier 2 / Prototype-capable
Synthetic data. No production APIs.
Tier 3 / UAT-connected
Scoped, time-bound access.
Tier 4 / Pattern lead
Engineering-owned release.
Evidence informs access.
Training develops capability. Access requires evidence, sign-off, and continued review.
- 01
Skills challenge
- 02
Partner review
- 03
Recorded placement
- 04
Renewal & appeal
03 / Research