Profiles become embeddings, embeddings become ranked introductions with a reason attached, and introductions become meetings on the schedule. Attendees stay in control of who sees them.
Networking is the reason people come; most apps leave it to chance
An alphabetical attendee list is not networking. Matchmaking is for events where the right five conversations are worth more than the keynote.
B2B conferences and summits
Delegates with a buying or partnering goal who need introductions to the right peers, not a crowded drinks reception.
Trade shows with hosted-buyer programmes
Buyers matched to exhibitors, meetings pre-booked at tables, minimum meeting counts enforced and reported.
Member and community events
Recurring communities where matching improves year on year because feedback is remembered.
How it works
From profile to meeting in five stages
The loop from meetings back to ranking is what makes day two better than day one. The privacy band is not a footnote; it decides whether attendees opt in at all.
Under the hood
Semantic similarity plus rules you can read
Profiles are embedded with the same approach we use for semantic search. Text from interests, role and goals becomes a vector, and the nearest vectors are candidates. An embedding captures that "procurement lead for hospital group" and "healthcare supply chain" belong together even though they share no words.
Similarity alone gives you people who are alike, which is not always who you want to meet. A rules layer re-ranks for complementary intent (buyers to sellers, hirers to candidates), role fit, mutual availability and exclusions. The weights are configured per event and visible to the organiser.
Every suggestion ships with a one-line reason: "You both listed battery storage and attend the grid-scale session". Attendees trust what they can read, and trust is what gets the meeting accepted.
Signal
Typical weight
Source
Declared interests and goals
High
Registration, app profile
Complementary intent (buy / sell)
High
Goal field, exhibitor catalogue
Role and seniority fit
Medium
Job title, group
Session favourites and attendance
Medium
App behaviour
Industry and company size
Medium
Registration
Past interactions and feedback
Rising over the event
Accept, decline, met
Availability and location
Filter
Agenda, meeting slots
Illustrative weights; the mix is tuned per event with the organiser.
What we deliver
Matching formats
Attendee to attendee
"People you should meet" in the app, refreshed daily, with one-tap meeting requests and digital card exchange.
Attendee to exhibitor
Stands ranked for each visitor from product catalogues; exhibitors see interested visitors and can invite them.
Hosted-buyer programmes
Pre-event matching, minimum meeting quotas, table assignment and attendance tracking for sponsors.
Meeting scheduling
Slots around the agenda, meeting points or tables, calendar invites, reminders and no-show handling.
Round tables and speed networking
Groups generated from the same embeddings, balanced for roles and companies, with rotations timed by the app.
Reporting
Suggestions shown, accepted, meetings held and feedback, per audience and per sponsor, in event analytics.
Process
From profile fields to day-two improvements
Design the profile
Which fields to collect, where, and how to keep it under a minute for attendees.
Configure rules
Intent pairs, exclusions, visibility defaults and the weights for this event's audience.
Dry run
Run matching on registered attendees a week out, review a sample with the organiser, adjust.
Launch
Suggestions go live in the app and by email before the event so meetings are booked on arrival.
Learn
Feedback re-ranks nightly; the post-event report shows what drove accepted meetings.
What goes wrong when done badly
Matching on job title alone, so every CFO meets every CFO. Black-box suggestions nobody trusts. Opt-out buried in settings, so attendees feel watched. Suggestions without a way to book, so interest never becomes a meeting. Each of these is a design decision, not a model limitation.
How does AI matchmaking actually decide who to introduce?
Each attendee and exhibitor profile is turned into an embedding from their declared interests, role, goals and the sessions they favourite. Similar embeddings are candidates; rules then adjust the ranking for role fit, buyer and seller intent, availability and exclusions. Each suggestion carries a short reason, and accepting or declining it feeds the next round.
What data does it need from attendees?
A role, an organisation, a handful of interests and a goal such as buying, selling, hiring or learning. Registration already captures most of it; the rest is a 30-second step in the app. Matchmaking works with thin profiles but gets noticeably better with goals.
Is it privacy-friendly?
Attendees opt in, choose who can see them, and can hide from exhibitors or specific companies. We never use sensitive attributes, and the reason for each match is shown in plain language. Profiles can be deleted after the event. See AI and data privacy.
Does it work for exhibitor-to-attendee matching and hosted-buyer programmes?
Yes. Exhibitor product catalogues are embedded alongside attendee goals, so a buyer sees the five stands worth visiting and a stand sees the buyers worth inviting. Hosted-buyer rules such as minimum meetings and table assignment are handled by the scheduler.
Can it run inside our existing event app?
Yes. Matchmaking is an API with suggestions, meeting slots and feedback endpoints. It ships inside Eventify and can be integrated into another app or a web portal.
Ready to make AI work in production?
Tell us about your systems, your data and your constraints. We reply within one business day with a straight answer and a suggested next step.