Inside our build: A 150-person AI summit, browser-native

Earlier this year, an executive community engaged us to deliver a two-day generative-AI conference. 150 attendees, three languages, hybrid format. The brief was specific: no headset requirement, no app installs, full multilingual delivery, and a recap they could share with sponsors within 24 hours of the closing keynote.

This is what we built, what worked, and what we'd change next time.

The brief

The community runs an annual flagship event. Their previous virtual programs had used Zoom and one of the early metaverse platforms. Both had problems. Zoom delivered no networking. The metaverse platform required a Quest headset and lost half the audience to onboarding friction.

For this year's event, they wanted XR's spatial advantage without XR's accessibility tax.

The constraints we agreed on:

• Browser-native delivery on desktop and mobile, with optional VR for attendees who chose it
• Real-time delivery in English, Spanish, and Portuguese
• Networking spaces designed for executive interactions, not virtual avatars bumping into each other
• Live event production support across both days
• AI-generated session summaries within the platform, deliverable to sponsors and absent attendees

What we built

Three Verses, one event. A keynote stage Verse for plenary sessions. A networking lounge Verse for breaks. A breakout cluster of three smaller Verses for parallel tracks. All cross-linked so attendees moved between them with single clicks.

Multilingual delivery in real time. Live captions in EN/ES/PT on all stage content. Chat translation across the platform. Speaker audio in the original language with simultaneous interpretation overlaid for attendees who selected ES or PT in their profile.

Branded environments, not generic templates. The keynote stage carried the community's brand — colors, logos, sponsor placements. The networking lounge had visual themes for each session topic, so attendees could find groups discussing the topic they cared about without scrolling through a list.

Producer console for the run team. Stage transitions, speaker switching, audience Q&A queueing, live polls — all controlled by the event producers without needing platform engineers on call.

AI-generated recap. At the end of each session, the platform generated a structured summary — speaker, key points, audience questions, sentiment indicators. By Day 2 closing, the production team had drafted recaps for every session.

What worked

Browser-first eliminated onboarding friction. Zero attendees missed a session because of install issues. Zero IT tickets. Average time from registration link click to first session: 47 seconds.

Multilingual delivery scored higher than English-only events. Post-event survey: 92% of Spanish-speaking attendees said the live captions were sufficient for full participation. 88% of Portuguese-speaking attendees said the same.

Networking utilization was 4x higher than the customer's previous virtual events. The combination of branded lounge environments and topic-themed groupings drove sustained networking interaction, not the brief lobby visits that dominated the customer's prior platform.

The AI recap delivered on day-two close. Sponsors received polished session summaries within four hours of the final keynote. The production team didn't write them.

What we'd change

Pre-event onboarding. We assumed attendees would explore on their own. Most didn't. For future events, we'd build a 90-second onboarding Verse — a guided tour of the event environments — that runs from the first registration confirmation email.

Speaker headshots on avatars. We used generic avatars by default with the option to upload a photo. Adoption of the photo upload was low, which made networking conversations feel less personal than they could have. We'd default photo-upload to required for speakers and strongly encouraged for attendees.

More recordings, longer windows. Attendees who couldn't make a live session wanted recordings inside the Verse environment, not just video files. We delivered video files. Next time, we'd build replay-mode environments that let attendees walk through the recorded session inside the Verse.

What this means for your event

Most enterprise event teams are still treating XR as either an overkill experiment (full Unity build for a flagship) or a watered-down upgrade to video conferencing (rebrand a generic metaverse platform). Neither is what XR can be.

The right pattern is browser-native, multilingual, and event-team-operable. If your event team can't run the platform without a vendor on standby, you're going to have a bad time at scale.

If you're scoping a flagship event for 2026 or 2027, we'd be happy to walk through what we'd recommend.