In the September 25 New York Times, Dr. Rachael Bedard made a confession I recognized immediately. AI, she wrote, is making her a better doctor, but she worries it's making doctors-in-training worse.
I read it twice. Then I sat with it for a while because I live on both sides of that sentence.
I've practiced small-animal medicine for 24 years. I'm also the Chief Veterinary Officer at OpenVet, where I help build the clinical AI tool veterinarians use every day. So, when she worries about what these tools do to young clinicians, she's describing a risk I'm responsible for.
Here's my honest response. She's right, and the people building these tools should say so out loud.
The word that stopped me
Dr. Bedard quotes Dr. Adam Rodman of Harvard on "never-skilling." The idea is simple and a little chilling. A seasoned doctor who relies on AI might lose an edge (deskilling). A student who relies on AI from day one may never build an edge at all.
The evidence is starting to arrive. In a 2025 multicenter study in The Lancet Gastroenterology & Hepatology, endoscopists accustomed to AI assistance detected fewer precancerous growths when working without it; their detection rate fell from about 28 percent to about 22 percent. These were experienced doctors. A May 2026 paper in Nature Medicine identified a risk for trainees and added another: "mis-skilling," in which a learner trusts an AI error and files it away as fact.
That last one keeps me up at night. A wrong answer that sounds confident is the most dangerous kind of answer.
Why this hits veterinary medicine harder
Human medicine has a long runway. Residency, often followed by a fellowship, then years of supervised practice. A resident can lean on a tool at 2 a.m., and an attending will still check the work at 7.
Most of our graduates don't get that runway. The majority walk out of school and straight into general practice. Some are working an emergency shift alone within months. Our supervised struggle that builds judgment is shorter. It's often thinner, too, because mentorship is the first to slip in a short-staffed hospital.
We're adopting AI quickly. In a Digitail and AAHA survey of 1,730 veterinary professionals this summer, 83.7 percent said they use AI, up from 39.2 percent in 2024. Most of that use is for documentation. Yet the same survey found that about two-thirds of practices adopted AI without significantly changing how they work. That's a lot of new power with very few new guardrails.
So the question Dr. Bedard asks about medical students is one we need to ask about our new veterinary graduates today.
What I believe the builders owe you
Keeping AI away from young veterinarians won't work. They'll use it anyway (most already do), and a good tool can make a tired new grad safer at all hours of the day. So the job falls to us, the builders, to build tools that teach as they help.
Here's the standard I hold our team to. It's also the standard I'd ask every veterinarian to demand of any vendor, including us.
Show the reasoning. A list of differentials with no path trains a clinician to copy. Show the path, and you train a clinician to think.
Cite the source every time. A vet should be able to check it and disagree with it.
Say what it doesn't know. A tool that never admits doubt teaches its user to stop doubting, too.
Let the doctor go first and have full authority. The best learning happens when you commit to your own differential, then compare it.
Keep a veterinarian accountable for the output. At OpenVet, I audit cases on the platform myself, with help from boarded specialists. Some of what I find goes directly back to the team for a fix.
None of this is glamorous. It's slower and costs money. It's also the only version of this technology I'm willing to put my name to.
What I'd ask of everyone else
To veterinary schools: decide now when students first encounter these tools and teach them to question the output. The Nature Medicine authors suggest a sequence: build baseline skills without AI, teach students to calibrate their trust, and then introduce AI under supervision. That's a sound place to start.
To practicing owners and group leaders: if you bought an AI tool, you also bought a training responsibility. Pair every new grad with a mentor who asks one question. "What did you think before you asked the tool?"
To investors: ask the companies you fund how they measure the impact of their tool on a clinician's judgment. Minutes saved is the easy metric. It's rarely the one that matters most.
To new grads: use the tools. Then close the laptop and explain the case out loud to someone. If you can't, you haven't finished learning it yet.
The part that gives me hope
Like Dr. Bedard, I'm a better doctor with AI beside me. It can catch the drug interaction a veterinarian would have caught on a rested day. It pulls up the paper I half-remember. It can give back minutes to spend with the client in the room.
But it helps me because I spent 24 years developing the judgment it relies on. My job as a CVO is to ensure the next generation has the chance to build that judgment, too.
Dr. Bedard ended on a note of worry. I'll respond with a promise from one builder: we heard you, and we're building for the student, too.
Sources
Bedard R. "Being a Doctor Will Never Be the Same After A.I." The New York Times, Sept. 25, 2026. https://www.nytimes.com/2026/09/25/opinion/ai-doctor-medical-students.html
Budzyń K, et al. "Endoscopist deskilling risk after exposure to artificial intelligence in colonoscopy: a multicentre, observational study." The Lancet Gastroenterology & Hepatology, 2025. https://www.thelancet.com/journals/langas/article/PIIS2468-1253(25)00133-5/abstract
Ke Y, et al. "AI-induced never-skilling in medical education." Nature Medicine, May 2026. https://www.nature.com/articles/s41591-026-04438-y
Digitail and AAHA. 2026 AI in Veterinary Medicine study (1,730 respondents, fielded July to August 2026). https://www.prnewswire.com/news-releases/veterinary-ai-use-reaches-83-7-and-practices-that-adopt-it-strategically-report-six-times-the-business-outcomes--digitail-and-aaha-study-302895402.html