Dwellci AI: AI-Assisted Architectural Drawings
Designed the MVP for Dwellci AI, empowering architects with fast idea generations and iterations through AI calculation and chats. Beta testing showed improved early development efficiency by 21%. MVP presented and helped secure pre-seed funding of Dwellci AI.
Dwellci AI is an agile startup that develops a digital architecture design platform empowered by AI. Instead of traditional early idea sketching and modeling process, Dwellci AI generates floor plans and 3D views within seconds, through direct AI chats. As the sole designer on the team, I designed and developed the Dwellci AI MVP.
COMPANY
ROLE
UX Design, UX Research, Product Strategy, Visual Design
3 Founders
TEAM
TIMELINE
(4 months)
Jun 2024 - Oct 2024
SKILLS / TOOLS
User Flows, Experience Mapping, Interaction Design, Figma
Stop sketching. AI-powered architecture starts here.
THE CORE EXPERIENCE
01 / USER RESEARCH
Architects juggle with sketching, revisions, and environment constraints in early idea development.
Through 2 user interviews and desktop research on architects’ journey in the early development phase, I was able to conclude 3 main challenges they face that could potentially benefit from AI.
Key findings from user research include:
Architects struggle to balance fast ideation with accuracy. Architects spend spending excessive time on early sketches that may require significant rework. They can benefit from higher efficiency.
They labor over syncing design views. There are multiple views of one architecture drawing, and when one view changes, architects need to manually update the rest of the views.
They need to comply with rule constraints. Architectural creativity must align with zoning laws, environmental regulations, and structural feasibility, making early-stage design both complex and restrictive.
The 4 stages of AI-assisted architecture drawings
Based on architects’ workflows and current pain points, I created an experience map of AI architecture drawings in the early development stage, generating insights in the experience structure.
02 / IDEATION
Try creating a new project study and design with AI.
In the ideation phase, I explored different ideas for 4 main features: home page, create a new study, comparing AI generated drawings, and optimizing design details.
03 / INFORMATION ARCHITECTURE
How do users start from project to canvas, and how do they navigate in the AI canvas?
Main user flow:
Describes the journey from homepage to design canvas, the main journey of focus.
I explored different navigation structure, and decided on this one to ensure users can chat and interact with AI smoothly, and pull up information they need, without displaying too much information to reduce cognitive load.
User navigation in the design canvas:
Intuitive AI interactions, combined with familiar tools.
04 / CONCEPT DEVELOPMENT
In the concept development phase, I explored more details about AI interactions, and how to make the experience intuitive for architects who never tried AI architecture drawings before.
I spend a lot of time exploring the best formats of AI generated design options, and how users can apply the options to the canvas. AI options should be clear, easily differentiated and understood, resulting in my design choice of option cards with images, title, and short descriptions.
AI compares and explains design rationales:
Users can track and reuse history drawings.
A specific use case in architects’ development process is they revise and iterate on their drafts a lot. How might we make the edit process painless and easy to track? I came up with the “version model” concept.
Users are guided step by step through study creations.
The primary design principle in this study is to make the AI experience as familiar and intuitive enough to reduce learning cost. I broke down user tasks into finest details to help users follow the process painlessly.
05 / END EXPERIENCE & IMPACTS
Imagine Julia’s daily experience with Dwellci AI.
Julia is working on a new academic building envisioned for Ethereal Muse Museum.
She is guided to input ground information that AI can further build on.
Julia asked AI to generate some initial ideas for her.
AI generates some quick drawings for her, including 3d block models, floor plan, and other views.
She is not super happy with the new version and wants to switch to a previous model.
She has the choice to deselect the current model and re-select a past generated model stored in history.
She thought the daylight setting is not optimal right now and wants to adjust it.
AI can adjust the design on a specific requirement or constraint.
Operational efficiency increased, and users are happy.
decrease in time spent on sketches and revisions.
22%
faster in design generation time: reduce from 60 min → 15 min.
75%
86%
customer satisfaction ratings in beta testing.
The End. Redirecting you to…
Funsport Education
Pathloom Travel
PrintCraft Hub