Insurers are enabling owners to have their pets checked out for potential health problems through artificial intelligence (AI)-based services provided through specialist veterinary apps.
The pet insurance industry is big business. According to MarketsandMarkets Research, the global pet insurance sector is worth about $14bn in 2025 and is projected to reach almost $30bn by 2030.
Rob Gray, head of engineering at Vet-AI, tells Computer Weekly how the firm is automating a part of a veterinary surgeon’s role using AI, reducing insurance costs as a result. The venture capital-backed company has been going for around six years.
Vet-AI’s main business comes through insurers, who offer customers a subscription to its service as part of insurance policies. It also offers a pay-as-you-go model direct to pet owners, but Gray said its primary customers are insurers. Customers will have unlimited access to vets via chat and telemedicine, as well as access to AI tools which automates “an element of veterinary care”, said Gray.
“Our primary mission is to deliver cheaper, more affordable and better clinical outcomes for pets by automating an element of the typical journey,” said Gray. “We’re able to save claims costs for the insurers and, if we do that, it drives down the price of insurance for the owners.”
The AI models will give advice on whether the owner can care for the pet at home, whether they need to see a vet – which Vet-AI can do through telemedicine – or whether it is best to go into a clinic. Vet-AI has 130 staff, with more than 60 qualified vets and 25 tech professionals.
The service is delivered via a mobile app and the web. To maintain and keep these up to date, Vet-AI automates testing and uses a continuous delivery model, with around 12 software updates per day. It currently has half a million pets registered in the UK through four insurers.
The AI bit
Around five years ago, the company built a decision tree symptom checker before it really got going with AI, but now uses bespoke in-house models – for example, it has one for gait analysis, which analyses the way dogs walk to detect abnormalities. This modal has been trained on around 20,000 dog videos collected at trade shows, clinical environments and the Vet-AI app. The owner simply videos the dog walking and AI will analyse it using this model.
Vet-AI also has a skin lesion model which is trained on about 80,000 skin lesions, all tagged by its veterinary team to detect dermatological conditions in pets from a photograph.
Then there is a triage model, which uses some of the base Gemini models from Google alongside retrieval augmented generation based on the half-million consultations Vet-AI has done.
“As it’s going through a conversation about your animal, it’s always referring back to the conversations vets have had with real-world clinical scenarios,” said Gray.
He added that the company has about 81% accuracy for its diagnoses where it uses VeNom Codes, alongside its models.
The diagnosis will then move to a model which decides the best form of treatment, whether online or in person. Because of possible AI hallucinations, and the element of risk, a human vet will get involved at the point of giving advice, which is much faster after many of the right questions have already been asked.
Vet-AI currently completes around 300 consultations per day.
