A dentist I work under recently started using an AI tool to help draft patient communication: preappointment instructions, followup texts, that kind of thing. Nothing clinical, just the soft admin layer around visits. From a marketing angle it actually works pretty well. The copy is cleaner than what we were sending before.
But something about it sits a little odd. Dental care is one of those contexts where patients are already anxious, and the language you use to reach them matters in ways that are hard to quantify. The warmth has to feel real or people notice, even if they can't articulate why. The AI output reads fine so far, but it's a bit frictionless in a way I can't fully pin down.
The question I keep coming back to is whether these models are actually getting better at contextsensitive tone, or whether we're just getting better at accepting outputs that are close enough. Those are different things, and I think it matters which one is true, especially in fields where trust is part of the product.
Curious if anyone here works in a service context, healthcare, therapy, legal, whatever, where they've noticed the tone gap narrowing or staying stubbornly wide. My sample size is small, ymmv.
[link] [comments]