Back to all work
Internal buildMVP in development

Building an AI-Native Social Planning Platform

Why social coordination is broken, and how AI assists group formation, venue discovery, and event planning.

Type
Internal build
Status
MVP in development
Tags
AI · Product engineering · Social
/ 01

Problem

Most consumer apps optimize for discovery (events) or messaging (chat), but planning real-world activities with strangers or shifting groups breaks down at the coordination layer. Time slots, venues, group sizes, and interests don't align — and group chats decay before plans land.

/ 02

Challenge

How do you assist (rather than automate) group formation and venue selection without removing user agency? And how do you bootstrap a network where the first user has no group to join?

/ 03

Proposed solution

An AI-assisted planning surface that suggests groups, venues, and times based on stated interests and a lightweight social graph. The AI proposes — humans confirm. Onboarding seeds activity templates so a first-time user can join an existing pattern, not an empty grid.

/ 04

Technology used

TypeScriptNext.jsPostgreSQLVector embeddingsLLM planning agentMapping APIs
/ 05

Business value

Not yet measured. Working hypothesis: the planning layer is the conversion bottleneck — moving event completion rate from low single digits to double digits is the success bar.

/ 06

Current status

MVP in development. Limited beta planned for 2026 Q3 with two interest-led communities (sports, local meetups) as initial verticals.

/ 07

Lessons learned

In social coordination, AI-assisted beats AI-driven. Users will let the system suggest, but always want final say. Discovery (more events) is not the problem — coordination (this event, this group) is. Internal research finding.

/ 08

Future roadmap

Group rating and reputation, venue partnerships with availability APIs, recurring-activity templates, lightweight membership tiers.

Have a problem that looks like one of these?