THE STORY

Defining an AI-assisted Architectural Concept Development Experience.

PRODUCT

AI SaaS Web App

◍ 0-1 MVP Design

◍ Shipped

ROLE

Product Designer

TEAM

2 Founders (Business +Tech Communications)

TIMELINE

Jun - Oct 2024

THE PRODUCT SPACE

Automating the Early Conception Phase of Architecture Design

Architects juggle between a number of softwares and metrics files during their early conception phase and are troubled with low efficiency and repetitive work. Dwellci AI aims to provide an integrated automation solution for their early exploration workflow.

MY PROCESS

Design the MVP from 0 to 1, realizing AI drawing, optimization, and export.

My 0 to 1 process to design the MVP can be categorized into 3 key stages, in each of which I solved for different sets of challenges.

STAGE 1 / KEY QUESTION

What challenges architects face, and what does the optimal experience look like?

UNDERSTANDING USERS

What are architects’ current workflows, and the challenges?

Micheal Reyes

  • Senior Architect & Studio Partner

  • 15 years in architecture — worked on residential, urban and civic buildings

Limited Exploration

  • They juggle to keep track of multiple design versions & decisions

  • Limited time to explore a lot of different options

  • It’s hard to ideate while fitting into site constraints

  • Explore more design options while remaining efficient

  • Keeping track of multiple design versions easily

  • Incorporate constraint check smoothly

Repetitive Actions

  • They have to bounce between multiple softwares and files simultaneously

  • Current design tools lack intent and there is no back and forth flows

The Optimal Experience

  • Address all the information in one place / transition easily

  • Input design preferences, intent

  • Edit design with back and forth discussions, like with real human

IDENTIFYING VALUE SCOPE

Software Collision

  • Drawings lose details when they export to other platforms

  • The data and visual models are asynchronous and both require manual updates

  • Export to other softwares with original fidelity

  • Update in data reflects on automatic update of views

What functions generate the most values for architects?

With the user insights, I conducted brainstorming and design analysis.

Key Concepts

The prioritization matrix results in 3 feasible and high impact features after discussion with the founder.

STAGE 2 / KEY QUESTION

What is the start to end experience? Are they addressing main pain points?

IDEA VALIDATION

Internal Testing: are flows intuitive, feasible, and meeting product goals?

With preliminary concepts, I created major flows and conducted internal testing with the founder to “color dot“ quick wins and potential risks with scope and technical constraints.

Study Creation Flows Internal Testing

Quick Wins ●

  1. The flow is intuitive and the architecture is efficient

  2. The embedded map is feasible

  3. The input metric is feasible

Potential Risks ● ●

  1. Limit functionality on parcels and generated models

  2. There is a wait time of 10-15 mins for AI generation

  3. There is a wait time of 3-5 mins for views change

AI Optimization Key Flows Internal Testing

Quick Wins ●

  1. The metrics are necessary for architects’ workflows
    The view change is intuitive and feasible

  2. Natural language input to AI is feasible and transfers to model generation

  3. Download is essential and feasible, in some way

IDEA DEVELOPMENT

Potential Risks ● ●

  1. Limit functionality to no manual editing on model/data for now

  2. Generating a new change takes 5-10 mins

  3. Integrating with other softwares are hard to realize

How might we optimize design based on testing insights?

1 / Scalable for multiple parcels, optimized for one.

Considering the MVP feature scope, I optimized the flow to demonstrate 1 parcel study creation. However, the creation structure can scale to multiple parcel and larger project creations in the future.

2 / Add manual data input and direct file type downloads.

To accommodate different forms of data input and model output, I incorporated manual data input form, and allowed simple model exports by directly downloading different file types to transition in other softwares.

3 / Limit AI generated options and provide predicted time.

Considering the backend generation of time of models, and the scope of MVP, I scoped down the design generations to 3, optimization to 1, and added estimated time indicator to educate users.

STAGE 3 / KEY QUESTION

Are the interfaces intuitive and straightforward?

VISUAL & INTERACTIONS

Intuitive, guided, and usable interactions.

In the final design stages, I iterated and designed affordances and guided navigation for intuitive experiences.

4 / Easily track metrics and update views during design

5 / Export different file types and seamlessly integrate with software

Progress Indicators

AI Explanations

KEY FEATURES

4.88 / 5 User Scoring in testing on “incorporating into current workflows!“

Layered Interactions

Generating efficient AI architecture development experience with 5 key features.

80%+ Task Completion Rates and Usability Scoring in testing!

1 / One time, efficient site creation

2 / Seamless data input and AI generation

3 / Chat with AI and get personalized edits

The End. Redirecting you to…

Funsport Kiosk

Outdoor Navigation

Craft Community