Dana calls Overhead Door Co. at 6:40 on a Tuesday because her garage door won't open and her car is trapped inside. No one picks up.

Thirty seconds later, her phone buzzes: 'Hey! Sorry we missed you, how can we help?' The tone is warm. The timing is not.

Dana knows a person didn't hear her voicemail, think about it, and type that in half a minute. She replies, 'my opener is dead and I need someone tonight,' half testing whether anyone is actually there. Nothing specific comes back. She calls the next company on the list.

What Dana felt has a name: the uncanny valley of trust. A reply that sounds robotic and arrives instantly gets a pass, because nobody expects a machine to sound like a friend. A reply that sounds like a friend and arrives instantly gets punished, because a friend who read her situation in thirty seconds is not a friend. It is a script wearing a smile.

Researchers at Tilburg University found this directly: making a support chatbot sound more human, without giving it more to actually say, lowered how much people trusted it. The warmth wasn't the problem. The mismatch between warmth and evidence of attention was.

Speed reads as proof of inattention, not proof of responsiveness. Generic phrasing confirms it. Dana doesn't need the text to feel like a person. She needs it to prove someone, or something, actually looked at her call before replying.

The fix isn't a better sentence. It's a system that waits before it writes.

missed call comes in → check the call time and the last service page the caller viewed → hold for a believable pause → send a text that names the specific detail, not just the greeting.

For Dana, that means a text at 6:41, not 6:40:30, that reads: 'Sorry we missed your call about the garage door opener. We're open until 8, want us to send someone tonight?' No exclamation point performing friendliness. Just proof the message came from her call, not a template.

A 2024 study on AI agents found that people forgive a slightly stilted or eerie delivery when the substance underneath is competent and specific. The pause and the reference do the work a receptionist's memory used to do, and they cost nothing extra to send. The pause is short enough that Dana never notices it as a delay, only as a text that finally answers her.

Webspenser · Reply to this email with the moment you'd like covered next.

Three things from the blog worth your time this fortnight.

Unchecked AI agents for business risk fast, costly errors. Map your automation boundaries and put technical kill switches in place before systems fail.

Stop sending robotic text replies to missed calls. Use our prompt frameworks and edit rules to build a human AI receptionist for small business.

PMC (Building User Trust in AI Chatbots for Customer Servic… · Explains how context-aware, personalized responses build trust while generic ones trigger suspicion.

Why people in your business behave the way they do, and what technology can do about it. Bi-weekly, three minutes.