A new report from AI company Upfirst analyzed 450,702 inbound calls handled by 503 AI receptionists and found the industry is shifting from broad adoption to targeted performance optimization. For customer support teams, the numbers clarify which setup choices actually reduce hang-ups - and which ones don't move the needle.
What the data says about hang-up rates
Only about 13% of the difference in hang-up rates between businesses could be tied to AI settings like greetings, voice, or training. The rest of the variation came from factors outside the AI's control, with many callers hanging up within the first few seconds, before any interaction with the system. That means a high raw hang-up rate often reflects the phone number's reputation rather than the AI's performance.
Still, several specific practices did correlate with lower hang-ups. Greeting callers with the business name, ending the greeting with a question, and telling callers they are speaking to an AI assistant all helped. So did informing callers that the call may be recorded and enabling capabilities like appointment booking, call transfers, and text messaging. Customer support teams exploring these optimizations can find structured guidance in the AI Learning Path for Receptionists.
What made no measurable difference
The study found that several commonly debated choices had little impact once other factors were accounted for. The gender of the AI voice, the length of the greeting, and the style of the opening question all showed no measurable effect on hang-up rates. This suggests teams can stop debating these elements and focus on the factors that actually change customer behavior.
Disclosure builds trust, not resistance
Alfredo Salkeld, Co-founder of Upfirst, said, "The common worry is that admitting it's AI answering the phone will make people recoil. We found the opposite to be true. Disclosing it's an AI is associated with about 20 per cent lower hang-ups..... Your raw hang-up rate grades your phone number's reputation more than your receptionist. Judge the AI on the calls where someone actually speaks."
The finding challenges the instinct to hide AI involvement. For customer support teams, it means transparency during the greeting can be a practical lever for engagement, not a liability.
Why this matters for customer support teams
You can't control the hang-ups that happen before a caller engages, but you can control the experience once they do. The report points to a clear checklist: state the business name, ask a question early, disclose the AI, mention recording, and build in real workflow actions like booking and transfers. These are concrete, low-cost adjustments that make a measurable difference. As the space matures, the winners won't be the ones who adopted AI first - they'll be the ones who optimized it based on evidence, not intuition. For broader strategies in AI-driven service, the AI for Customer Support resource covers deployment and integration patterns across the entire support stack.
Your membership also unlocks:
