Veterinary AI meeting looks for clinical value in data

In terms of artificial intelligence, the veterinary profession has progressed from experiencing a data shortage to a shortage of usable outputs.

Miel Hostens, a professor with Cornell University’s College of Agriculture and Life Sciences, says, “Data generation is not an issue anymore. The challenge is what to do with the data.”

He gave a keynote talk at Cornell’s second Symposium on Artificial Intelligence in Veterinary Medicine (SAVY 2.0) May 16-18, which drew 130 attendees to Ithaca, New York, and more than 80 virtual participants from 23 countries. To broaden access, meeting organizers provided free virtual admission to participants from low- and middle-income countries.

SAVY 2.0 organizers presented the symposium, with the theme “Innovation, Inclusion, and Impact,” as a working forum to meet the veterinary profession’s near-term needs rather than as a showcase for speculative technologies. The two-day meeting featured more than 20 presentations on AI-related work in wildlife health, One Health surveillance, livestock analytics, and companion animal diagnostics.

Results pending

Hostens characterized the past “AI decade” as heavy on investment and high expectations but generating little in the way of practical value.

He described ongoing work in AI-supported precision dairy management practices that combine camera feeds, sensor streams, and health and production records to support herd health. Using computer imaging to detect subtle behaviors indicative of, say, the early stages of lameness, is progressing, according to Hostens, as is machine proficiency in analyzing radiographs. However, broadly generalizable technology developments remain limited, he added.

“The big success stories are not there yet,” Hostens said, noting that marketed capabilities such as lameness prediction still need to show consistent value on the farm. He emphasized that progress is closely tied to data custody and governance. Small animal hospital-controlled datasets, especially imaging, allow for standardized labeling and defined access. By contrast, production animal data are scattered across farms, laboratories, and third parties, with stronger privacy and commercial constraints.

To overcome these barriers, Hostens has applied for National Science Foundation funding to support his work using AI to collect and pool relevant data that omits potentially sensitive information. He also highlighted a project linking Journal of Dairy Science content with a chat-bot interface, allowing users to quickly retrieve research quickly rather than manually searching archives.

Hostens expects food animal veterinarians will one day use machines capable of prioritizing patients before the veterinarian arrives on site. He described a likely near-term workflow in which a practitioner asks, “Which clients need attention today?” and then receives a list of specific animals for examination on the basis of herd history, recent behaviors, and environmental data.

Such systems, Hostens said, could reduce the time it takes a new veterinarian to grow in confidence and their decision making.

Workflow efficiency

Former veterinary technician Allie Aspen spoke about how AI-powered documentation can return time, and some calm, to veterinary staff.

Aspen is currently a clinical solutions specialist with Scribenote, an AI-driven language processor for veterinary practices. She recounted watching veterinary colleagues drown in burnout, working 16-hour shifts, missing meals and time with family.

It got too much for Aspen, so she took a job as a medical scribe working in human medicine where she was responsible for transcribing doctor’s notes. She immediately noticed what a difference that made. “The doctors were present. They were connected with their patients. They were also leaving on time and able to spend time with their families,” Aspen said.

Eventually she returned to veterinary medicine and helped implement an AI transcription program for Veterinary Emergency Group. The result was less time spent on documentation and the updating of medical records, easing workloads and allowing for more attentive care. In her view, AI is less a novelty than a support for over-stretched staff.

AI transcription technology reduces time spent on medical records an average of 90 minutes, according to Aspen, even allowing for the addition of an extra case and a half to a veterinarian’s daily schedule. Those extra minutes turn into better days and lighter workloads, she said.

Aspen recounted a hearing-impaired veterinarian telling her that with documentation off his hands, he could connect through “facial expressions and body language” without worrying that notes would suffer. A clinician with severe anxiety wrote: “My anxiety about cases walking through the door turned to enthusiasm to see more cases and help more people and their pets … Instead of feeling overwhelmed by being busy, I felt energized by it,” she said.

“This isn’t about AI,” Aspen said. “It’s about building a future where care is sustainable to everyone who gives it.” For busy, short-staffed clinics, she characterized the calculation as practical: tools that shrink paperwork and restore presence help to serve patients and clients first and not the spreadsheet, and that, she said, is how efficiency becomes wellbeing.