Transforming an AI-Generated MVP into a Scalable Product
The evolution of an AI-generated MVP into a consistent product system.
B2B SaaS Web & Mobile
Year
2026
Role
Product Designer
4 weeks
Team
Design + Engineering
Challenge
Results
The AI-generated MVP was transformed into a structured product system with clear flow logic and consistent behavior across platforms.
Rebuilt the lead management flow based on real sales journeys.
Redefined the information hierarchy to reduce cognitive load in critical decision-making moments.
Replaced ambiguous interactions with clear system states and transitions.
Standardized logic across Web and Mobile, ensuring consistent behavior.
Consolidated fragmented flows into a single, unified lead operations system.
The result was a product where Web and Mobile shared the same structural logic, with Mobile-first execution as the primary priority.
50+
Screens Restructured
100%
Lead lifecycle mapped
1
Validated Workflow System
Trade-offs & Constraints
• Flows generated by Lovable could not be directly reconstructed in Figma: they required a full rebuild from scratch rather than a translation.
• The MVP had structural redundancy: duplicated flows and components solving the same problems in different ways.
• The company’s Design System only supported mobile. I used shadcn as a foundation for web, maintaining product logic consistency without relying on a shared DS.
• The leads dashboard required handling real user data (volume, errors, incomplete information), which the original fictional MVP had never accounted for.
• Several original prototype features (JSON reports, sketch generation, multiple webviews) proved to be high-cost and low-impact after refactoring.
As a result, the business deprioritized the web version and proceeded exclusively with Mobile.
Process
Heuristic Diagnosis of the AI-Generated MVP: I began by conducting a heuristic analysis of the AI-generated product, identifying structural issues such as:
Weak information hierarchy
Lack of consistency across interaction patterns
Cognitive overload in critical flows
Absence of clear system states
Poorly prioritized actions within interfaces
This analysis made it possible to separate UI problems from product logic issues.
Sales System Mapping: Before redesigning any interface, I structured the full lifecycle of a lead, including:
Sales user journeys
Lead states across the sales cycle
Qualification and progression criteria
Success, failure, and edge cases
State transition rules between stages
This process established a single source of truth for product behavior.
Learnings & Conclusion
This project demonstrated how AI tools are effective at accelerating early exploration, but insufficient for structuring product logic and complex UX decisions.
The main transformation was not visual, but structural: moving from a set of AI-generated screens to a coherent and scalable product system.
Key Learnings:
AI accelerates the starting point, but does not replace product thinking.
System architecture is more important than interface design.
Defining what not to build is a critical decision: even when it means recommending the cancellation of an entire platform.
Cross-platform consistency depends on logic, not UI.
Usage context defines experience architecture.