Building in Sri Lanka, from anywhere
Problem
Building in Sri Lanka from abroad runs on trust and presence, with nowhere to find verified experts.
Role
Designer and builder, end to end: product, marketing site, both apps and the design system.
Approach
One booking loop, an AI first responder and one design system for two roles, built with Claude Code.
Status
Built to MVP and live as a working prototype to explore.
Planr is a consultation platform for Sri Lankans living abroad, and overseas investors, who want to build or invest at home but can't manage it from the ground. It connects them with verified architects and construction specialists, and lets them book, follow and ask questions without being there.
The construction market runs on trust and physical presence. From overseas there was no single place to find verified experts, no way to check credentials remotely and no clear way to hire with confidence. For younger earners doing it for the first time, the barrier was higher still: the income and the intent, but no idea where to start.
Thinking
Four decisions, and what each one gave up
Every decision came back to one question: can someone make a confident decision about a building they can't visit?
Book from live availability
Not chosen: Request and replyChosen: Pick a live slot
Across time zones, every message round trip costs a day. Browsing profiles, picking from real availability and confirming in one step removes the back-and-forth entirely.
Trade-offIt only works if consultants keep their calendars current, so availability had to be quick to manage.
Verified experts, not an open marketplace
Not chosen: Open listingsChosen: Verified profiles
Overseas clients can't check credentials in person, and the market runs on trust. Every profile is verified, so choosing someone from abroad can be a confident decision.
Trade-offA slower start on the consultant side, since every expert has to be verified before listing.
AI as a first responder
Not chosen: AI as the advisorChosen: AI between sessions
Planr AI, powered by Claude Haiku, answers architecture, permit and cost questions in real time, positioned to help between sessions rather than replace the consultant.
Trade-offBigger decisions still wait for a human, by design, so the AI never has to sound more certain than it is.
One design system, two mental models
Not chosen: Two separate appsChosen: One shared system
Clients are in discovery mode and consultants are in management mode. Both are served from one component set, with layouts tuned to each role rather than separate products.
Trade-offComponents had to flex for both contexts, which took more care up front than designing each app alone.
Design
Two roles, one set of components
The client side is built for discovery and the consultant side for management. Both use the same cards, counts and actions, arranged for what each person is there to do.
Status
What the MVP covers
The full loop was designed and built with Claude Code and works end to end, and the prototype is still live to explore.
Marketing site
Explains the service to people deciding from abroad, and leads to early-access signups.
Client app
Browse verified profiles, book from live availability and follow each consultation.
Consultant app
Manage bookings, sessions and client questions from one dashboard.
Planr AI
Answers architecture, permit and cost questions between sessions, using Claude Haiku.