Kinney Drugs Pulls Back AI Pharmacy Assistant After Customer Complaints

What Went Wrong

The core problems weren’t subtle.

• Refill errors: Customers were approving AI-initiated refill requests that turned out to be inaccurate, resulting in unwanted prescriptions.
• Consent by default: Anyone who gave Kinney their phone number was automatically enrolled in AI interactions—no explicit agreement required.
• Privacy concerns: Experts raised questions about the handling of protected health information, which Kinney disputed but couldn’t fully defuse.
• Customer confusion: The AI was handling calls in a way that left patients uncertain whether they were talking to a person or a system.

The company’s president acknowledged publicly that the rollout didn’t get customer experience “right.” That’s a measured way of saying the gap between what the tool was designed to do and what patients actually experienced was significant.

The Consent Problem Is the Real Story

Burt was framed as a way to free up pharmacists for clinical work by handling routine interactions. That’s a reasonable goal. The execution, however, leaned on passive consent—customers were opted in simply by having a phone number on file.

In healthcare, that’s a high-stakes assumption. Patients dealing with chronic illness or complex medication schedules aren’t a forgiving test group for an AI still finding its footing.

One customer described repeatedly approving AI refill requests she assumed were accurate, only to receive medications she didn’t need. That’s not a UX friction problem. That’s a patient safety problem.

Where Regulation Stands

Healthcare AI deployment is moving faster than the rules designed to govern it. Vermont’s new data privacy law, signed in June, doesn’t take effect until 2028. That’s a long runway for tools handling sensitive health data to operate in a relatively thin regulatory environment.

Kinney isn’t alone in this. Healthcare systems across the country are adopting AI tools at a pace that legal frameworks haven’t caught up with. The Burt situation is a visible example of what that gap looks like in practice.

What This Means for AI Tool Decisions in Healthcare

The Kinney rollback is a useful case study, not just a cautionary tale.

A few things stand out for anyone evaluating AI tools in patient-facing or high-stakes environments:

• Default opt-in is a liability. In healthcare especially, consent should be explicit and easy to understand.
• Routine doesn’t mean low-risk. Prescription refills feel administrative, but errors have real consequences.
• Rollout speed matters. Deploying broadly before the tool is calibrated to your specific patient population compresses your margin for error to near zero.
• Customer trust is slow to build, fast to lose. Kinney’s customers noticed the shift away from human interaction immediately—and had strong feelings about it.

The company is now back to touch-tone phones for calls and opt-in texts for refills. That’s not a failure of AI as a category. It’s a failure of deployment strategy—and a reminder that “supplement the team” only works if the team, and the patients, are actually on board.