General interest: AI – be wary of the honeyed words

When I was relatively early in my career, the company for which I had worked for the past two years was sold to a large corporate enterprise. When the team found out, people were worried that our identity would be diluted, costs (and jobs) cut, and profits prioritised.

Two messengers from the buyer were dispatched. They wore smart casual outfits and were well groomed. They came from a veterinary background. They brought doughnuts. They assured us that our fears were misplaced, that they weren’t the big bad corporate overlords, that nothing would change. Much. That we would just have access to more efficient systems, lower medication costs, a greater collective culture of excellence.

Anybody who read the CMA’s findings can tell how that went.

I was reminded of this reading a recent article on artificial intelligence (AI) in the veterinary sphere: “If technology could lighten the load, even by a margin, isn’t it worth asking not why we would use it, but why we wouldn’t?” Well… you did ask.

There are excellent reasons to like the idea of AI in the veterinary world: tailored reminders for clients, accuracy of noting and dosing, none of that boring cytology (more on that later); the perfect angelic guide on your clinical shoulder – right next to where your stethoscope goes.

But let’s ask: why not?

Firstly, for that guiding presence: we have the study from MIT and Cornell universities which suggests that the employment of AI to serve our cognitive needs leads to their diminishing. Frequent users become less capable of solving their own problems, stuck in the honeyed words of large language models which can always serve you another clinical baklava – perfectly packaged and deliciously straightforward, but without any of the graft that makes such a meal meaningful¹.

If we stop writing notes entirely, and instead depend on the scribings of an AI model, the cognitive work performed in this action that embeds and sorts the information obtained is lost for good. Do I miss things out? Yes. Do I like taking notes? No. Do I need to order my scattered thoughts by writing and thinking? Absolutely.

You cannot learn to ride a bicycle by having a robot balance the bike and pedal you along. You’re not cycling; you are being cycled. If we give up on more and more of the work of veterinary medicine we will, over time, be reduced to greeters, little more than a smile and a reassuring word. Until “Tilly Norwood” takes over that job, too.

Next, as AIs become more complex and competent with blinding speed, our ability to discern the inner workings behind their decisions will become increasingly opaque, reducing our ability to tell a hallucination from an insight. “So, doctor, why did you give Fluffy Bunnikins Rabbit co-amoxicillin, iron tablets and pentobarbital up its nose?”

“VetGPT told me to,” is unlikely to suffice.

The Veterinary Surgeons Act is not even up to date with the demands of today’s world; how can it possibly cope with what comes next? Should something go wrong, it seems unlikely that the large multinational corporations which will purchase these systems and consolidate them are going to end up footing the bill.

Moreover, if their motivations are opaque, what stops them suggestion-bombing you into the use of proprietary products or services? Will these suggestions serve your patient? You? Them?

Currently, these products and services seem entirely benign, even miraculous. The issue is the speed of change when you have a technology that is not simply a tool, but is instead an active, iterative force.

So, what should we do?

Am I just screaming into the silicon void? The best advice I’ve heard (listen to the podcast, Your Undivided Attention with Tristan Harris – it’s brilliant and terrifying) is to run a pre-mortem.

Rather than waiting for the genie to leave the bottle with a whiff of brimstone and sector-changing consequences, leaving us to ask ourselves, “what happened there?”, ask yourself these questions now:

1. What are all the things that could go wrong with the implementation of this technology?
2. How do we ameliorate these risks?

Can we form groups of concerned professionals and lobby our professional bodies or the corporations who employ us for fair treatment?

Can we affirm that we won’t accept the laying off of the most vulnerable members of our teams in the face of this technology?

Can we discuss, dissect and determine the motivations and means of those selling these new systems?

Business leaders will tell you that AI is “nowhere near” replacing you. That may be true. But little by little, it will devour the work that for so many of us provides so much meaning. That would be a hell of a loss.

• This article appeared in Vet Times (1 September 2026), Volume 56, Issue 35, Page 11

Will Trewarman graduated from the University of Cambridge in 2014 working as a small animal vet and locum in England and Wales before settling in the West Midlands. He is in independent small animal practice working towards a dermatology certificate, with interests in communication, leadership, teaching, and the future of veterinary practice.