UI Bakery AI Agent
Role: Product Designer
Duration: Jul 2025 - Present
Team: 7 people, including 3 PMs, 4 Engineers, 1 Product Designer (me).
Domain: Web Application · SaaS · B2B AI · Low-Code / Developer Tools · Conversational UX
Long story short
In 2025, I led the design of UI Bakery’s AI Chat Agent - transforming the platform from a traditional low-code builder into an AI-powered environment where users can build full internal apps from a single prompt. As the sole designer, I shaped every layer of the experience: the conversational flow, prompt-to-app generation, motion feedback system, data integration, and design-system evolution for AI features. I also created ready-to-ship motion prototypes with Claude to make AI interactions feel seamless and alive. The result - a fast, trustworthy AI experience that boosted activation x10 compared to the main platform.
Context
After establishing UI Bakery as a mature low-code platform, we started exploring how AI could radically speed up internal app creation.
We took it step by step - and I contributed as a designer to several key AI-powered releases:
2023
UI Bakery AI Assistant
Chat-based helper inside the builder that supports debugging, explains code, and connects to documentation.
April 2024
UI Bakery AI App Generator (PH link)
The first version of the AI Agent that could generate internal tools and admin panels from SQL data.
October 2025
UI Bakery Agent for internal app generation
A reworked version of the agent that generates complete internal tools from a single prompt - with data connections, logic, and simple UI animations included.
The AI Chat Agent project started in mid-summer 2025, with a compact team of 4 engineers, 3 product manager, and myself as the only designer.
Our goal was to merge conversational AI with the builder experience - so that users could describe what they want, and the system would instantly generate a deployable app.
Context
After establishing UI Bakery as a mature low-code platform, we started exploring how AI could radically speed up internal app creation.
We took it step by step - and I contributed as a designer to several key AI-powered releases:
The AI Chat Agent project started in mid-summer 2025, with a compact team of 4 engineers, 3 product manager, and myself as the only designer.
Our goal was to merge conversational AI with the builder experience - so that users could describe what they want, and the system would instantly generate a deployable app.
! Problem
As UI Bakery evolved into an AI-first platform, we faced a key shift:
moving from a structured builder to an open-ended, prompt-driven experience.
This introduced several critical problems:
Low-code felt like building, not generating - users were forced into developer-like thinking.
AI didn’t replace the workflow, only assisted it - limiting real efficiency gains.
Delayed feedback and lack of structure reduced trust and slowed down time-to-value.
Design challenges
Bringing AI to production
Turned AI from demo into a real app builder with data, logic, and clean UI.
Designing AI experience
Used motion and micro-interactions to make AI feel fast and clear.
System & ownership
Updated the design system and owned the product end-to-end as a solo designer.
Solutions
What was done and what benefits were achieved for each solution:
Prompt-to-App
A conversational interface as an industry standard now where the user describes their app (“I need an order management dashboard connected to my PostgreSQL data”).
👉 AI generates the complete UI, structure, and logic in ~2 minutes (depending on request).
Smart visual & motion feedback
Each step of the process was visualized through smooth motion cues that made AI generation feel natural and alive.
👉 I used Claude to create purposeful animations and deliver ready-to-use motion prototypes, which developers could easily copy and integrate - removing adaptation issues and simnifically speeding up delivery of this small features.
Data integration & Data storage
Apps can be generated using user-provided or mock data. In this iteration I created new flows of how
👉 Users can switch sources easily via the native UI Bakery database or their own connectors or securely connect to third-party tools.
Workflow generation, Code preview & export
The system automatically builds triggers, actions, and conditions in the workflow editor. You can switch to Code view to review, edit, or export the generated logic and files directly.
👉 There was a decision of adoption of previous already built solutions that reduced time-to-market.
Fast start & templates
Designed a flow from a landing page with engaging animations to organiztion with ready-to-use templates, and AI input where users can type their first prompt and choose a starting point right after login. They land directly in the builder, ready to connect data and start generating.
👉 Cut time-to-value to minutes, letting users start building meaningful apps within seconds.
Research, Discovery & Validation
AI Chat Agent was rolled up in early October 2025.
For validation, I used different technics:
Customer feedback
Got regular feedback from users through support and product analytics.
Behavioral analytics
Used PostHog to watch sessions and understand user behavior.
User testing
Ran user tests, quick corridor tests, and customer interviews.
Impact & Results
“Wow! I tried building apps with UI Bakery AI. What the hell! That’s really mind-blowing! Works so good… Impressive! Good job!”
UI Bakery enterprise client, used previously just low-code platform.
Business outcomes
Higher activation in AI Agent vs. low-code product.
Up to 15× faster app delivery.
Improved conversion and retention across the funnel.
Strong lead-to-activation growth.
Positioned product as an AI-driven platform.
(Accurate numbers are under NDA)
Design leadership impact
Led end-to-end design of AI Agent (UX, flows, motion, feedback)
Turned AI delays into smooth, predictable interactions
Scaled design system for AI-generated UI
Maintained high quality at speed (~86% UX satisfaction)
Reflections and Takeaways
Drove UI Bakery’s evolution from a theme tool to an AI-powered platform where teams build apps from a single prompt - achieving several times higher activation than the low-code product.
The AI Chat Agent became a key milestone, turning the platform into a conversational AI tool that generates full apps in minutes.
I designed the end-to-end UX flow - from the first prompt to generation and feedback - keeping the process clear, interactive, and trustworthy.
Reflection
This project proved that conversational UX is not about chat, but about behavior, timing, and trust.
While the low-code builder required long, predictable flows, here the challenge was to make a short interaction feel alive and confident.
Designing with AI became about turning user intent into real outcomes through clarity and responsive feedback.
