NailyAI:
AI-powered Design Tool for In-house Designers
NailyAI is an internal AI-powered design tool built for in-house nail designers. Existing AI tools kept producing visually creative outputs that couldn't survive real production constraints -- so I repositioned the product around production feasibility rather than creative novelty, embedding material and brand requirements directly into the generation process.
Product Manager
My Role
1 CEO
1 Product Manager & Designer (Me)
1 Visual Designer
3 Developers
Team
6 Months
Duration
User Research
AI Product Strategy
UX Flows
High-Fidelity Prototyping
Key Contribution
01 Real-world Impact
4.8/5
Satisfaction Score
Usability Study
36 designers
21%
Design Efficiency
Before: 190 / week
Now: 230 / week
20x
Asset Search Speed
Before: 6 mannual steps, 1min
Now: 1 click, 3s
02 Problem Reframing
Initial Direction — Boost Creativity to Increase Output
We worked with a press-on nail design and production company whose business depends on both creative design and scalable manufacturing. Their in-house nail designers are responsible for creating large volumes of design concepts that later move into production, where consistency, feasibility, and efficiency are critical.
The client approached us with a clear goal: to increase design output by introducing AI into their designers’ workflow. They believed that by “boosting creativity” with AI, designers could generate more ideas, faster — ultimately improving productivity and speed to market.
Research Lens — Aligning Business Goals with User Workflows
Before introducing AI into the workflow, I needed a clear picture of how nail designs were created day to day — including where time was spent and what constraints shaped production decisions. To gain this visibility, I structured the research around two lenses (business constraints and designer workflows) following the initial stakeholder meeting, and collaborated with a teammate to execute it over two weeks.
Business
User
To learn about the problem scope, business impact, and operational constraints.
To understand current workflow, user pain points & needs, AI tool opportunities.
Stakeholder Meetings
(n=3)
Field Research
(n=2)
User Interview
(n=3)
Survey
(n=36)
Key Insight — Creativity Without Constraints Didn’t Translate to Productivity
Through analyzing research findings, I identified that designers were actually not struggling to generate ideas, and that existing AI tools were already capable of producing visually creative outputs. The real breakdown occurred when these outputs failed to survive real-world production constraints — rarely translating into usable design references or measurable productivity gains.
Therefore, I framed this gap between creative generation and production usability as the core problem NailyAI was designed to address.
Original Assumption → Superficial
“If AI helps designers generate more ideas faster, productivity will naturally increase.”
Reframed Understanding
“Productivity is not driven by how many ideas AI can generate, but by whether creative outputs can survive real production constraints and function as usable design references within daily workflows.”
03 Product Strategy & Positioning
Core Strategy — From Model Parameters to User Intent
As a product designer working on AI-powered tools, a core challenge was translating complex model parameters into controls that designers could understand, trust, and use confidently. To address this, I worked closely with engineers to align model capabilities with designers’ mental models, and validated design decisions through iterative user testing.
AI models operate through detailed, low-level controls.
AI Ability
Nail art designers think in style, materials, techniques, and overall vibe.
Designer Mental Models
NailyAI
What designers can control
What the model generates
The Bridge:
Conceptual Grouping
Mapped low-level model parameters to design-oriented controls.
Designed interactions around design intent, not technical parameters.
Intent-First Interaction
Constraint-Based Controls
Used constrained ranges and presets to reduce unusable outputs.
Product Positioning – A Domain-Specific AI for Nail Production
To scale effectively, we made a deliberate positioning choice: NailyAI would not compete as a general AI image tool, but serve nail production workflows where feasibility and execution constraints matter.
Scope Choice
We intentionally narrowed the scope to nail production.
Instead of targeting broad creative use cases, we focused on a domain with clear material constraints and production standards—allowing AI outputs to be evaluated against real-world feasibility.
Advantage Source
Our advantage came from embedding production knowledge.
By incorporating materials, techniques, and brand requirements into prompt logic, we shifted from generating “pretty images” to generating executable designs.
Value Proposition
We optimized for production efficiency, not visual novelty.
The goal was faster decision-making and higher reuse alignment—reducing iteration waste and improving workflow throughput.
05 Design Focus 2 — Explore: Supporting Inspiration & Reuse
Create Flow Prototype
Core Challenge — Beyond a Chatbox, Without a Parameter Wall
📍 The core challenge was finding the middle ground between two extremes.
A chatbox-only experience made generation feel easy, but produced inconsistent results that designers couldn’t confidently evaluate or reuse. On the other hand, exposing low-level model parameters increased control, but created a technical UI that didn’t match designers’ mental models.
I needed to reframe “AI generation” into a design step—where constraints are intentional, understandable, and still fast to iterate.
Design Approach — From Analysis to Structure
To avoid designing in a vacuum, I first ran a lightweight competitive analysis across inspiration, creation, and asset-management tools to understand common patterns and gaps.
With this context, I mapped the Create experience through information architecture and flow diagrams to clarify where decisions are made, how designers and AI share control, and where constraints should guide generation.
Information Archtecture:
Core User Flow:
Direct competitors:
Adjacent competitors:
Key Design Decisions — Where Constraints Live
Designing the Create experience required more than exposing AI capabilities. The core challenge was deciding what needed to be constrained early, what should remain flexible, and how designers could make confident decisions without managing technical complexity.
Because nail designers evaluate through visual comparison, the layout prioritizes the generated result as the visual center. Keeping outputs in focus supports faster judgment, while grouping controls on the side enables iteration without breaking visual attention.
1.Nail Setup — Defining Structural Constraints Up Front
Instead of relying on prompts to describe structural decisions, I treated nail setup as a foundational step that defines non-negotiable constraints early—such as nail shape and color structure—so generation starts within a feasible design space.
2.Model Setting — Shaping Output Behavior, Not Exposing Parameters
In early iterations, I intentionally explored multiple model settings to understand how much control designers actually needed. Through user testing, I found that additional controls increased cognitive load without improving decision-making.
Based on this, I deliberately reduced these into a single, designer-facing control that reflected the variation designers actually cared about, removing parameters that added noise rather than value.
① Designer Control (Visual Slider)
② Product Representation (Tags)
3.Prompt Input — Making Design Intent Explicit Before Generation
Early prompt input often consisted of fragmented keywords or vague descriptions, they rarely translated into reliable outputs—making it hard to tell whether unexpected results were caused by model behavior or unclear intent.
To address this, I introduced a “Refine with AI” step that restructures rough input into a clearer, reviewable description. This allowed designers to verify intent alignment before committing to generation, improving both output predictability and confidence.
① Guided Prompt Placeholder
② Designer Draft (Keywords)
③ AI-Refined Intent (Reviewable)
04 Design Focus 1 — Create: Turning Constraints into Creative Control
Explore Flow Prototype
Core Challenge — When Inspiration Isn’t Reusable
📍 The core challenge was that design assets were abundant but functionally unusable.
Within the internal design library alone, over 5,000 high-quality nail designs were stored. However, most assets were only searchable by filename or upload date — not by visual attributes, production constraints, or design intent. As a result, designers relied heavily on memory or manual browsing to find references, limiting reuse and slowing iteration.
I needed to transform inspiration from static images into structured, searchable design references.
Design Approach — Translating Engineering Outputs into Design Structure
I realized that inspiration images were treated as static visuals, while designers actually needed design references — assets that carry structure, intent, and production relevance. Without that structure, even the best images remained hard to retrieve, compare, or reuse.
While the underlying system handled data processing and metadata extraction, my role was to define the design abstraction: what information should surface to designers, how it should be grouped, and how inspiration could function as a reusable building block rather than a one-off reference.
Raw Image
✅ Visual appearance
Design Asset
✅ Visual appearance
✅ Style
✅ Structure
✅ Characteristics
Reusable Reference
✅ Visual appearance
✅ Structured meaning
✅ Searchable
✅ Comparable
✅ Reusable in workflows
Key Design Decisions — Designing a Continuous Exploration Loop
1.Search Bar — Structuring Filters to Guide Exploration
At the entry point, my goal was to make searching feel intuitive rather than technical. I structured the search experience into two distinct layers: resource scope and design attributes, allowing designers to narrow intent without committing too early.
Resource Type
Catergories
Based on production constrains
Based on current asset management approach
2.Design Detail Card — Making Inspiration Understandable
Browsing alone isn’t enough for reuse. Designers need to understand why a design works, not just what it looks like. The design detail card surfaces structured attributes alongside the original visual. This turns each image from a passive reference into an interpretable design asset.
3.Apply to Create — Carrying Exploration Forward Into Action
Exploration becomes truly valuable only when it feeds directly into creation. “Apply to Create” allows designers to carry selected attributes and intent from Explore into the Create experience, instead of starting from a blank state. This preserves design momentum and reduces redundant decision-making.
