AI for
real life.

Research on human connection in the physical world. We build the AI that decides who should meet, invites them, runs the room and follows up. Companies, conferences and communities use it today.

Scroll
Mission

The internet connected everyone. Then everyone stayed home.

The decisions that shape careers, companies and communities are still made in person. Yet the physical world has had almost no intelligence built for it. Our mission is to build AI that understands people where it matters most: who should meet, where, how, and what should happen next.

Research

Four intelligences. One gathering.

Every gathering moves through four stages: who comes, who meets, the room they meet in, and what lasts. We build one intelligence for each: a model, the product people use, and the data that makes the model better, from our research and our own apps. A customer’s data stays with that customer, and we never train on it.

  1. 01ParticipationBefore
  2. 02ConnectionMeeting
  3. 03ExperienceThe room
  4. 04OutcomeAfter
01 · Participation Intelligence

Bring the right people together.

Matches each potential participant with the right inviter, message, and channel, personalizing the journey from interest to arrival.

Research
Why one person says yes to an invitation and another never answers: who asks, what they say, where, and when.
Model
Invitation policyChooses the inviter, message, channel and moment for each person, and is retrained on every response.
Product
Invitation agentsPersonal invitations that reach people where they already are: email, LinkedIn, X and text. Each one goes out in a real inviter’s name, and only with their approval.
Data
Every open, reply, decline and arrival, across every inviter and channel.→ makes Invitation policy better
Inviter → message → yes → arrival
02 · Connection Intelligence

Make every interaction matter.

Identifies mutual value between people and shapes the format, duration, and focus of their interactions.

Research
Who should meet whom, and why: the value two people can create for each other, measured in both directions.
Model
Mutual valueScores every possible pair across a gathering, then plans who meets whom, when, in what format, and about what.
Product
Guide app and agentA mobile and web app with a conversational agent that guides each person before and through the event: who to meet, why it matters, and where to find them.
Data
What people tell the agent they want, who they actually meet, and how each meeting went.→ makes Mutual value better
Connection Intelligence · reasoning

03 · Experience Intelligence

Create the conditions for connection.

Adapts the setting, activities, and flow to participants’ interests, needs, and goals.

Research
How space, sound, seating and timing change what people say to each other.
Model
Venue simulationSimulates the room before the event; during it, agents read the room and decide what to change.
Product
Intelligent venueOpt-in cameras, displays and connected apps that run the room: a tested plan before the doors open, then seating, prompts and timing adjusted as the evening unfolds.
Data
How every room moves: where people gather, how long conversations last, which layouts work. Raw footage is deleted after the event.→ makes Venue simulation better
Small tables
04 · Outcome Intelligence

Turn experiences into lasting value.

Guides follow-through, measures outcomes, and learns from each experience to improve the next.

Research
What a gathering changed for each person: the relationships that formed, the work that followed, and whether they felt they belonged.
Model
Outcome modelAgents follow up with each person, turn what they hear into a measure of what the gathering was worth, and train the other three intelligences on it.
Product
Personal gifts and feedbackAfter the event, each person receives a gift chosen for them and a personal follow-up that asks what mattered. They feel heard, and their next invitation, match and room are shaped by what they said.
Data
Personal feedback from every guest, and what happened in the weeks after: second meetings, relationships and work.→ makes all four intelligences better
The room ends. The relationships don’t.
Deployed

Built in the lab. Proven in real rooms.

Our technology already runs the full loop, from the first invitation to the follow-through, wherever people meet: on campus, at conferences and around the table.

7,000+people matched to one another, in person
Each dot is ten people
Universities

Stanford and beyond

Matching and gatherings for graduates, from the first introduction to the people they keep meeting after.

Conferences

HumanX

Attendees matched, prepared and brought together at one of the leading AI conferences, with follow-through after the last session.

Companies

Cerebras, Zoom, Zscaler

Dinners for their customers, buyers and partners, from 20 guests to 300.

Every deployment ran all four intelligences. Each customer’s data stayed in its own system

Notes

From the lab. What real rooms teach us.

Short notes on the problems behind each intelligence, and how we approach them.

01 · Connection

Mutual value beats similarity

Most matching systems pair people who look alike: the same title, the same industry, the same interests. In real rooms, the conversations that lead somewhere usually connect people who need each other. Connection Intelligence scores both directions of a meeting, what each person can solve for the other, and makes an introduction only when the value is mutual. It then chooses the format that fits: a one-to-one, a shared table, or a small circle after a talk.

02 · Participation

The inviter is part of the invitation

Whether someone comes depends on who asks as much as on what is offered. The same gathering gets a different answer from a stranger, a colleague or a friend, and over email, a text or a direct message. Participation Intelligence treats the invitation as a policy to learn: for each person, which inviter, message and channel give them a real reason to say yes, and what helps them actually arrive. Every message still goes out only with the inviter’s approval.

03 · Experience

Seating is an optimization problem

A dinner of 20 and a dinner of 300 ask the same question at different scales: who sits next to whom. Experience Intelligence treats the room as a set of constraints, tables, seats, sound, time and flow, and the matches as the value to maximize. It plans and simulates the layout before the event and adapts the format on the day: small tables when people need depth, one long table when the group should become one, circles after a talk.

04 · Outcome

Measuring what a gathering changed

Who came says little about what changed. Outcome Intelligence follows what happened after: which introductions led to a second meeting, which relationships kept going, which work followed. Those outcomes are the training signal for the other three intelligences, so who is invited, who is matched and how the room is set all improve with every gathering. Real life, not time on a screen, is the measure.

Who it’s for

We build with the people who run the rooms.

01 · Partners

Field marketers

Their own field marketing agency, run by our team and agents. They sell executive dinners in their own name; we run them. They keep 25% of every contract.

Become a partner → 02 · Partners

Recruiters

They sell research dinners with the people a company most wants to hire, in their own name; we run them. They keep 25% of every contract.

Become a partner → 03

Companies

Buyers, customers, your industry and your next hires, at one table. Your first dinner in three weeks.

For companies → 04

Conferences

Every attendee matched with the right people, side dinners run, and follow-up after, alongside your event app.

For conferences → 05

Communities

Members who actually meet, in every city, without your organizers doing it all by hand.

For communities →
Principles

Technology for people, in the company of people.

Our work touches who people meet, what they are invited to, and what is known about them. These are the commitments we hold ourselves to, in the research and in every product.

01

Mutual or nothing

We introduce two people only when both of them gain. What someone privately wants is never shown to the other side; a match appears only when the interest is mutual.

02

Invite, never manipulate

Every invitation carries a real reason from a real person. No fake scarcity, no false urgency, and anyone can say no and be left alone.

03

Nothing sent without approval

Agents draft; people decide. No message goes out in someone’s name until that person has approved it.

04

Perception by consent

Face check-in and presence sensing are opt-in, used only for the event they were given for, and deleted after it.

05

Private by design

Every person’s and every company’s agents run on their own isolated computer in the cloud, with their own accounts. Not even we can open it, and we never train on a customer’s data.

06

Real life is the measure

We count relationships formed and work that followed, never time spent on a screen. The best session ends with people in a room.

Work with us

Let’s build something real.

Tell us about the gathering, the community or the system you want to build.

[email protected] →