Kinney Drugs pulls back AI phone assistant after hundreds of customer complaints

A regional U.S. pharmacy chain has withdrawn an AI-powered phone assistant after hundreds of complaints about wrong dosages, missed prescription notifications and general frustration reaching a human. Commenters argue that voice-based AI is technically brittle—especially with drug names, accents and complex edge cases—and that using it as a gatekeeper in healthcare is dangerous when prescriptions can be life-critical. Others note pharmacies are under intense financial pressure and see automation as a survival tactic, but many customers view these systems as cost-cutting that degrades care and erodes trust.

AI Phone Assistants in Pharmacies

  • Many see pharmacy phone lines as a terrible first target for voice AI: older callers, complex meds, insurance issues, low tolerance for failure.
  • Some argue only simple tasks (store hours, basic refills) should be automated; “refill exceptions” and edge cases should go straight to humans.
  • Others note pharmacies have long used rigid IVR systems; AI is just an attempt to expand automation beyond those narrow flows, with mixed or negative results.

Customer Experience and Frustrations

  • Multiple anecdotes describe AI systems that:
    • Fail to understand non-standard problems.
    • Loop users through scripted “self-service” answers already tried.
    • Make escalation to humans difficult or require yelling/abuse keywords.
  • Callers often prefer long hold times with humans over fast but unreliable AI, especially for medications or banking.

Technical and Implementation Challenges

  • Practitioners say voice AI is still “really hard,” especially:
    • ASR (speech recognition) error rates, worse with drug names and regional accents.
    • Limited context windows and brittle prompt setups leading to inconsistent behavior.
  • Effective deployments are described as requiring deep domain expertise, pharmacist project managers, and tight scoping to transactional tasks.
  • Others counter that real-world systems often use weak, cheap models and shallow integrations, making them worse than basic phone trees.

Healthcare Risk and Ethics

  • Strong pushback on “rough around the edges” in healthcare: edge cases matter because “people die.”
  • Concern that AI adds an “accountability sink”: patients can’t “ask for the AI’s manager,” while line staff absorb blame for systemic decisions.
  • Some fear AI will gatekeep access to care or meds, amplifying existing insurance and pharmacy bureaucracy.

Economics and Incentives

  • One side: retail pharmacies are financially squeezed (PBMs, insurers, low margins) and need automation to survive and reduce hold times.
  • Other side: AI cuts labor costs and boosts profits while degrading care; pharmacies and AI vendors are happy even if patients are not.

Broader AI CX Discussion

  • Comparisons to offshoring and kiosks: looks efficient on paper, erodes service and brand long term.
  • General pattern: AI works well for narrow, simple tasks; for complex, emotional, or exception-heavy interactions it often fails and increases anger.