How a Founder Conference Got 9 in 10 Through a 37-Min Survey

The chair of a national conference for a global peer network of business owners wanted each of a couple hundred attendees to meet the person they most needed to meet. The default was an alphabetical seating chart. Doing better required detailed attendee data, which meant a long survey with every question required.

North Group built an 80-question survey in 13 sections, every question required, with two consent questions at the end. A custom AI model read each submission and paired attendees for small-group sessions, one-to-one meetings, breakfast tables and mentorships, using three pairing criteria set by the chair and guardrails that kept existing groups and couples apart and mixed chapters and tenure. An AI agent wrote a personal readout of about 800 words for each attendee who opted in, delivered by email with a designed PDF. Pairings were revealed on site through sealed envelopes, name cards and table assignments.

The chair set three pairing criteria up front, and North Group added guardrails so existing groups and couples stayed apart and chapters and tenure got mixed. The survey was designed long on purpose: 80 questions across 13 sections, all required, so the model had enough signal to match well. Two consent questions at the end controlled whether an attendee received a readout. As each submission came in, the matching model scored it against the rest of the pool and placed the attendee into small-group sessions, one-to-one meetings, breakfast tables and mentorships. A separate AI agent then wrote a personal readout of roughly 800 words, delivered by email with a designed PDF, usually within about 15 minutes of submission. At the event, pairings were revealed in person through sealed envelopes, name cards and table assignments.

Conference organizers, peer networks and associations that want attendees to meet the right people and have the attendee data, or the willingness to collect it, to do it properly.