Beyond the Buzz: How AI Agents Transform Group and Event Sales in Hospitality

Felix Undeutsch, CEO and co-founder of hivr.ai, delivered a keynote at ITB Berlin 2026 demonstrating how agentic AI is transforming the meetings, groups, and events (MICE) sales process in hospitality. The talk opened with a compelling illustration of accelerating technology cycles: AI-generated video quality improved dramatically from 2023 to 2025/26, with Undeutsch having personally prompted a video of someone eating pasta at Berlin's TV Tower and flying over the Reichstag overnight — work that would have required professional production teams just a few years prior.

Undeutsch framed the broader context by charting how long it took major companies to reach $100 million in revenue: Siemens took 80 years, IBM 34 years, Microsoft 10 years, LinkedIn 7 years, Facebook 4 years, HN 2.5 years, Cursor 1.8 years, and Lovable just 8 months. This acceleration, he argued, is driven by AI product cycles that have no discrete end — each output feeds the next training loop, compounding improvement continuously.

He identified three macro shifts for knowledge workers in 2026: (1) AI usage is moving from optional to mandatory — employers will require it; (2) usage is shifting from individual/ad-hoc to organization-wide, with multiple agents collaborating across teams; and (3) the focus is shifting from efficiency (generating content faster) to measurable bottom-line and top-line impact — CFOs are now demanding ROI.

The core of the presentation focused on the MICE sales problem: the process is highly manual, driven by long email threads, versioned PDFs ('version final final'), and human data-entry between demand-side platforms (e.g., Cvent, Outlook) and supply-side hotel systems (sales & catering systems, PMS, CRM). hivr.ai deploys agents in the middle to act on data flowing between these two worlds.

Three live agent demos were shown in real time (timestamps visible in the videos, no acceleration):

1. Lead Qualification Agent: A meeting planner email was sent for 150 guests for a 3-day event in Las Vegas. Within 30–60 seconds, an AI voice agent called the planner's mobile phone, confirmed event dates ('November 15th arrival, meetings on Monday'), asked about date flexibility (none in this case), explored destination openness (Las Vegas preferred, open to other US locations), and gathered all qualifying information — a process that typically takes hotel sales staff hours or days.

2. Lead Scoring Agent: A simpler email (22-person meeting, August 28, accommodation needed) was sent. In real time, the system executed 6–8 automated tasks — reading/classifying the email, identifying the hotel, parsing the request, creating a lead, extracting data — and scored the lead at approximately $34,000 in deal value. The scoring incorporated multiple dimensions: completeness, availability, demand levels, occupancy context, wash risk, cancellation risk, and no-show costs. No human touched the process.

3. Proposal Agent: A complex multi-day event request (General Electric, November 15–17, accommodation + meetings + F&B) was emailed to the hotel. Within approximately 60 seconds, the system created a deal in the hotel's internal system, matched all prices and inventory items, structured functions by day and time, generated a branded client-facing proposal (with images, room descriptions, allocated meeting spaces), pulled the company name from the email signature, and delivered the proposal directly to the planner's inbox ready for e-signature.

Speed was presented as the decisive conversion lever: proposals sent on day one convert at roughly double the rate of those sent on day two.

Real-world results were presented from two named customers: - Flemings Hotels (Frankfurt): Tested H1 2024 vs. H1 2025. Lead-to-sale conversion rate went from 9% to 18% — nearly doubling. Revenue also nearly doubled (with a note that smaller meetings, which convert at higher rates, meant revenue growth was slightly less than 2x). - Durrint Lintner (unnamed property): Response rate to incoming leads improved from 52% in 2022 to 99% post-automation. The previous failure was attributed to COVID-era staff shortages and furloughs, which left half of all requests unanswered.

In Q&A, Undeutsch confirmed that the three demos were a subset of possible automations. Other covered use cases include: automated follow-up and nudging, AI-driven negotiation, handling Cvent's 20+ standardized RFP questions, rooming list management, mapping guest names to rooms, importing names directly into PMS, and shuttle scheduling based on flight arrival data. He also gave a shout-out to competing/complementary startup Onai as doing similar voice AI work in the hospitality space, encouraging the audience to explore multiple vendors.