AI Workflow Assistant
Deciding when AI acts, and when it doesn’t
The Challenge
Business Goals: Define a scalable AI vision without slowing delivery
Establish a clear, future-facing vision for how AI fits into the ad tech platform.
Enable rapid experimentation and delivery of AI features without blocking active engineering work.
Build a system that can adapt to new AI capabilities as the technology evolves.
Design Strategy Goals: Clarity, alignment, and consistency in a fast-moving space
Align leadership, product, and engineering through a shared design vision.
Create processes that kept the team current on emerging AI patterns and technologies.
Maintain design cohesion and quality despite short timelines and continuous change.
Research
Foundational Research
Industry Landscape — Reviewed articles, research papers, and presentations to understand current trends and emerging practices in AI.
Competitive Analysis — Analyzed 7 competitors' approaches to designing and integrating AI into their products.
User Needs — Surveyed 131 participants and held 15 in-depth interviews to identify user expectations, pain points, and opportunities for AI.
User Expectations
Context-aware responses
Interactive back-and-forth guidance
Validation before applying changes
User Concerns
Task errors/inaccuracy
High manual effort for error correction
Lack of transparency
Developing Design Principles
Affinity Diagram: Shaping Core AI Design Principles
Synthesize Insights — Combined user feedback, competitor analysis, research, and internal expertise to capture what makes AI design effective.
Identify Priorities — Organized themes and voted to surface the most critical principles that would guide the product.
Translate to Action — Turned each principle into tangible design guidance, ensuring consistency and measurable impact across the platform.
Design Principles
High-Quality, Reliable Output
Ensure output is aligned to user intent
No AI for AI's sake — Only apply AI when it adds real value
Continuously refine outputs using user input and contextual signals
Accountable for Errors
Set clear expectations for what the tool can and cannot do
Highlight potential consequences, especially in high-risk situations
Provide ways for users to recover or refine results when errors occur
Save Users' Time
Ensure AI workflows are faster and easier than manual alternatives
Provide clear, intuitive guidance for using the tool
Integrate seamlessly into existing workflows and minimize wait times
The Human is in Charge
Design AI to assist, not replace, the user
Require user permission or confirmation before taking action
Respect user preferences and avoid interrupting workflows
Auditing Existing Work
Heuristic Evaluation: Reviewing and Refining Early AI Features
Gather and Review — Collected all screens from in-progress alpha features into a single Figma file for easy review and comparison.
Assess Against Principles — Rated each feature in a shared Google Sheet to determine how well current designs aligned with established design principles.
Iterate and Build Library — Revisited designs that fell short to meet design principles and added components to a living library for consistent reuse in future designs.
Product Vision
AI Opportunity Mapping: Cross-functional workshop to drive strategy alignment
Defined Opportunity Space — Leveraged prior design research to identify user touchpoints where AI could add real impact, such as troubleshooting and optimization.
Cross-functional workshop — I brought together Product, Engineering, Design and Research to generate ideas, align on goals, and ensure all voices are heard.
Impact-Driven Prioritization — Breakout sessions: Engineering assessed feasibility, Design/Product assessed user value. We plotted outcomes on a complexity vs. value matrix.
Outcome: Short-Term Wins, Long-Term Vision
Rapid Impact — Selected low-complexity, high-value opportunities to deliver a compelling demo within three months.
Shape Strategic Roadmap — Leverage high-value, complex opportunities to position our product ahead of competitors and shape future AI roadmap.
Design System
Living Library: Ensuring design consistency during fast-paced iteration
Component Library — Designed components just-in-time to maximize speed, keeping the team updated through frequent communication to ensure consistent reuse.
Workflow Examples — Documented use of components in context to communicate how they function in practice.
Organizational Impact
Established a Common Language
Developed AI design principles and patterns that became the North Star for design and product teams, enabling rapid alignment on decisions.
Ensured User Trust
Applied a user-centered approach to feature ideation, addressing early concerns about reliability and intrusiveness to create AI tools users actually embrace.
Future-Proofed the Design Process
Built a technology-agnostic design system that supports rapid iteration and accommodates emerging technologies, keeping teams aligned as designs evolve.
Enter password from resume to view project
© 2026 Eugenia Lee