AI voice agents for hotels: lessons from 15,910 real guest calls (polydom.ai)

A recent analysis of 15,910 incoming calls handled by AI voice agents across a European hotel network revealed critical insights into guest behavior and the pitfalls of traditional phone systems. When the hotels removed their Interactive Voice Response (IVR) menus, which prompted callers to select their preferred language, the empty call rate unexpectedly increased from 12% to 19.5%. This was not a failure of the AI technology; rather, the IVR had been silently filtering out callers who couldn't navigate the menu, effectively hiding thousands of lost conversations. The analysis highlighted that differing hotel types—spa vs. city hotels—faced unique challenges, with city hotels experiencing a spike in dropped calls due to their international clientele. To address these issues, the hotel group is testing a bilingual greeting that provides a welcoming message in both the local language and English, thereby eliminating the IVR but maintaining clear communication. This approach aims to improve the caller experience and reduce the hidden loss of potential bookings. Key takeaways for the AI/ML community include reassessing the impact of existing systems on data interpretation and the importance of immediate caller engagement, particularly for international guests. As the hospitality sector increasingly adopts AI, understanding the nuances in data can significantly affect operational success and revenue.