Episode 16: AI Receptionists and the Next Security Wave

Dr. Adam Little, Robert Sanchez, and Dr. Aaron Massecar return with a fast-moving roundup of the biggest veterinary AI stories. The episode centers on a lawsuit involving Zoetis and a misdiagnosed cancer case, then expands into the practical realities of AI receptionists, scribe integrations, and what comes next for governance and security.In this episode, the hosts focus on how the profession should judge AI systems: by perfection, or by relative risk compared to current human-led workflows. They also dig into how voice AI, integrations, and agentic tools are changing daily practice, while warning that security and validation are now urgent priorities.

Key topics

• The Columbia Veterinary Hospital lawsuit against Zoetis, where an AI-assisted screening result allegedly contributed to a missed cancer diagnosis, repeat surgery, and the death of an 11-year-old dog.
• Whether the legal and ethical blame sits with the AI company, the veterinarian, the hospital, or the broader workflow around the case.
• Why the hosts keep returning to relative risk, not perfect performance, when evaluating AI in healthcare and veterinary medicine.
• The concern that companies may market AI with "shock and awe" claims while downplaying limits, failure modes, and validation data.
• How this case could influence veterinarian adoption by increasing caution, liability concerns, or skepticism toward AI-assisted pathology and radiology.
• The rise of AI receptionists and why a 25-minute failed booking experience for a Bordetella vaccine appointment matters as a client-experience case study.
• The idea that bad outcomes often come from workflow failures, not just model failures, especially when latency, context, or agency breaks down.
• Recent veterinary software movement toward deeper integrations, including two-way connections between scribes and PIMS systems, plus MCP-style access to platform data.
• Why integrations matter beyond copy-paste convenience: fewer windows, less friction, better focus, and more of the practice happening in one place.
• The next big concern after integration: security, including sandbox escape risks, hardened validation, and compliance language like ISO 27001 and SOC 2 Type II.

Timestamps

00:00 - Welcome back and what the veterinary AI news cycle has been missing
01:03 - Columbia Veterinary Hospital lawsuit and the Zoetis AI misdiagnosis claim
02:55 - Who is responsible when an AI-assisted decision goes wrong
04:21 - The veterinarian's responsibility versus specialist or lab guidance
06:13 - Was the product positioned clearly as a screening tool
07:35 - Why the hosts reject "is it perfect?" and focus on relative risk
09:14 - Why this lawsuit feels like a broader blame-diffusion problem
10:29 - Why this case is messy compared with a more direct AI failure
12:22 - Could this change veterinarian adoption of AI tools
14:19 - Why edge cases will multiply as more practices use AI daily
16:42 - Why AI literacy and fluency matter more than blind trust
19:01 - Workflow failures and the bridge to AI receptionists
20:27 - The Bordetella booking story and why client friction drives clinic switching
22:26 - Human call friction data and why status quo is not great either
24:20 - Why older AI phone experiences created lasting skepticism
25:41 - AI receptionists, client AI calls, and the demand for transactional workflows
28:53 - GPT Live and the leap in real-time voice interaction
31:43 - Three ingredients of a good phone experience: latency, context, agency
35:59 - New integrations with PIMS and AI tools, including two-way and MCP-style connections
38:46 - Why integration reduces app-switching and improves focus
41:27 - "Tool use" and the future of agents acting on behalf of the practice
44:38 - Why the first scribe workflows had a useful human-in-the-loop bug
46:32 - Validation as building the road, not just the car
47:28 - Security threats, sandbox escape, and why this is the next urgent infrastructure problem
51:35 - How much AI has already replaced in everyday work, from setup to form filling
53:30 - Final takeaways on adoption, implementation, and staying close to the frontier