An AI-powered voice assistant can search, respond, adapt, and finalize a booking or order within a single conversation. This was demonstrated by Pieter Peeters (GTM), Boning Gong (FDE), and Shana Inch (GTM) during an ElevenLabs webinar, using an agent built with ElevenAgents. The speakers showed how this agent handles real hotel booking conversations.
Customers don’t move in a straight line online. They change their dates or budget, ask questions along the way, and expect the answer to lead them to the next step. Forms and filters often force them to start over. According to ElevenLabs, 70% of shopping carts, payment funnels, and online bookings are abandoned.
During the demo, a concierge agent installed on a booking site helped a customer find a hotel in Amsterdam for two people, under €400 per night, within walking distance of the canals and the museum. The customer extended her stay from two to three nights during the conversation. The agent spotted an overlap with a prior booking in Istanbul and asked for confirmation before proceeding. It also asked the customer about a birthday and passed a note to the hotel. The booking was confirmed, and the conflicting reservation was canceled upon request. Finally, the summary was provided in Dutch, and the conversation then switched back to English.
Results Published Across Several Sectors
ElevenAgents is used in hospitality, travel, retail, commerce, and marketplaces. At SevenRooms, the Voice AI assistant has handled over 400,000 calls for more than 600 restaurants, with 1 in 5 calls occurring outside business hours. According to SevenRooms data, comparing the first three months of 2026 to the same period a year earlier, restaurants saw up to 25% more bookings.
At Rohlik Group, one of Europe’s largest online grocery players, the Maia agent handles 90% of incoming calls to the support line and 90% of customer interactions via phone, web, app, and WhatsApp, in six languages. According to the company, request resolution is more than twice as fast, and the group plans to reduce its operational costs by 50% while maintaining its customer satisfaction score (CSAT). “We believe voice is the future of grocery, ” says Lukas Honke, Head of AI Engineering at Rohlik Group. At Klarna, the agent handles the first level of phone support for all of its 35 million US customers.
Other cases are cited: at Immobiliare.it, Italy’s leading real estate platform, lead qualification improved by 3.3 times, and users were 74% more likely to share their phone number. At KPN, the Dutch verification rate rose from around 50% to 90% through joint work.
Why Voice, and Why Configuration Matters
In a study conducted by ElevenLabs with 98 people, 85% preferred voice over chat for rebooking a canceled flight. According to the same source, beyond two minutes of conversation, voice succeeds 29% more often than chat. The data also mentions latency, meaning the agent’s response delay: a 16% decrease in this delay was accompanied by a 26% increase in the success rate.
Configuration remains crucial. According to ElevenLabs, many disappointing agents are poorly configured, whether it’s the model, the voice, or the time spent tuning them. Teams that test for only a few hours sometimes conclude the technology isn’t ready. To build an agent on their own, teams can purchase a license, while forward deployed engineers or trained partners can assist them. According to the company, teams that build entirely on their own only leverage 10 to 20% of the platform’s potential.
