The HIMSS Conference Nobody Actually Attended
Abstract
The conference floor at HIMSS is already kind of a simulation. Anyone who has walked the Las Vegas Convention Center in March, badge scanning through vendor booths while someone in a logo polo tries to schedule a “brief 30-minute follow-up,” knows the theater of it. The whole thing is performative sales infrastructure dressed up as knowledge transfer. Which makes the following thought experiment less crazy than it sounds: what if the humans just stopped going, and sent their AI agents instead?
This is not a piece about the metaverse. That hype cycle is mostly dead and nobody is pouring another drink for it. This is about something that is actually being built right now, across a dozen enterprise software stacks simultaneously, and the question is just whether the health tech industry is going to be intentional about it or stumble into it backwards.
Key Arguments:
- The primary value of conferences like HIMSS is top-of-funnel discovery and partnership exploration, not the content sessions
- AI agents are already capable of executing the semantic matching, negotiation scaffolding, and follow-up workflows that conferences are supposed to accelerate
- A new business model exists for conference operators who want to survive the transition rather than be bypassed by it
- The technical architecture is largely available today using existing LLM infrastructure, RAG systems, and multi-agent orchestration frameworks
- The human role shifts from “attendee” to “deal closer” with significant productivity implications for enterprise health tech sales teams
Data Points Referenced: HIMSS 2024 attendance (~28,000 registered), average enterprise sales cycle in health IT (12-18 months), cost per attended lead at major health tech conferences ($800-$2,000+ all-in per meeting), LLM context window capabilities relevant to negotiation tasks.
Table of Contents
The Conference as a Broken Sales Primitive
What AI Agents Can Actually Do Today
The Architecture of an Agent-to-Agent Conference
Business Models for the Operator and the Attendee
The Human Closes the Deal
Why HIMSS Should Build This Before Someone Else Does
The Conference as a Broken Sales Primitive
Here is what HIMSS actually is if you strip away the educational programming, the keynotes, the Harold Klieman Foundation awards, and the slightly depressing buffet situation: it is a very expensive, very slow, very geographically concentrated CRM event. Companies spend north of $100,000 just to have a booth presence when you factor in the exhibit space, shipping, booth build, hotel rooms, flights, meals, and the opportunity cost of pulling your best people off their actual jobs for a week. In return, they get badge scans, some conversations that were already scheduled two months in advance anyway, and a pile of business cards that someone will fail to enter into Salesforce before February is over.
The content sessions are largely a sideshow. The real work of HIMSS happens in the Starbucks line, the suite meetings at the Wynn, and the private dinners that the companies with real budgets host on Tuesday night. The conference is functioning as a temporal and geographic forcing function. It gets the right people in the same place at the same time, which is genuinely valuable, but the mechanism is preposterously inefficient relative to the outcome.
To be specific about the numbers: HIMSS 2024 drew approximately 28,000 registered attendees. Of those, a meaningful fraction are there in explicitly commercial roles, meaning they are either selling, buying, or evaluating partnerships. If you assume conservatively that the average enterprise health tech company spends $150,000 to attend, and that they walk away with 40 qualified conversations, the cost per meaningful interaction runs somewhere between $800 and $2,000 when you account for the ones that go nowhere. That is an extraordinarily high price for what is essentially a discovery mechanism, a way to find out that a company you had never heard of has a data asset or a workflow capability that maps onto a problem you are actively trying to solve.
The tragedy is not that conferences exist. The tragedy is that the discovery layer, the part where you figure out who you should be talking to and why, is done almost entirely by humans wandering around a floor plan with a printed map, hoping they bump into the right booth. In what other domain would you accept that level of signal-to-noise for that price?
What AI Agents Can Actually Do Today
