The strongest pattern in the comments was not "AI bad" so much as "this use case is backwards." Plenty of people said narrow automation already works. Refill IVRs, stock checks, appointment prep by
SMS, and similar transactional flows can be useful when the system stays inside a tight lane. The failure comes when companies stretch that into open-ended voice support for healthcare, where the long tail dominates and the consequences are real. Several practitioners said production voice systems still break on speech recognition, accents, and domain vocabulary like drug names. Others pointed out that even if the underlying model is decent, most companies cheap out on weaker models and brittle orchestration, then present the result as if it were a competent front desk.
A second thread sharpened why these rollouts feel hostile. The complaint is less that machines make mistakes than that companies remove the correction path. AI gets inserted as a gate in front of the person who used to absorb nuance, use judgment, and waive small frictions. In banking, telecom, retail, and pharmacy alike, commenters described bots that cannot admit failure, keep reciting the obvious, and waste twenty minutes before allowing escalation. In healthcare that lands harder because the issue is often not convenience but access to medication. Even people open to automation kept coming back to one rule: if intent is unclear or the request falls outside a well-tested script, apologize and transfer immediately, with transcript in hand.
There was one credible pro-AI case, from operators building pharmacy voice tools, and even that case reinforced the same lesson. They argued the technology can work when it is implemented with deep pharmacy expertise, conservative scope, and serious operational tuning, not as a generic "voice AI for X" product sold by outsiders. That did not change the overall verdict. The problem here was not that callers failed to appreciate a promising technology. It was that a cost-cutting system was dropped into a high-trust workflow before it was dependable enough, and customers noticed immediately.
wcax.com