Every guest at a 60-person dinner got three personal introductions that read like the host wrote them

The problem

A private members club runs a heavy calendar of salons and founder dinners. The touch that makes the evening is a personal note from the host: "I want you to meet these three people." Doing that for 60 guests requires knowing every attendee's work, interests, and history, and no host has that for a room this size. The club's member data existed, but split across an events base and a 13,000-person community database that were only joined by email.

What we built

• The system joins event RSVPs to the 13,000-person community database, pulls each confirmed guest's bio, interests, and background, and flags anyone whose profile is too thin for a good introduction so a human can fill the gap before the event.
• Each guest gets matched to three other attendees on shared interests and industry resonance, with guardrails that read like a senior host's instincts: skip family members, skip colleagues, cap how often any one person gets recommended so nobody appears in fifteen emails.
• Every introduction is a third-person one-liner under 120 characters, written to be read in an email from the host. Thirty in a row still read like a person wrote them.
• The host reviews and tweaks every note before it goes out. The AI never hits send. That is the difference between scaling intimacy and cheapening it.
• A companion bio book ships with each event: a private, members-only site with every guest's photo and background, publishable for a new event in under an hour.

Stack