Since its release to the public in November 2022, ChatGPT (and other artificial intelligence tools) have become commonly accepted by the public as a place to go and ask questions, as well as somewhere to produce colouring pages for the kids and redesign our living rooms. As a vet, I’ve definitely seen the shift from clients having asked ‘Dr Google’ before they seek my opinion to now having asked an AI instead (with varying levels of accuracy and safety). 

AI is here to stay and we’re all using it… so is your vet using it too? A recent survey showed around 21% of vets are regularly using AI in their working lives, and this number is set to rise as practices incorporate the use of veterinary specific AI programmes into their computer systems. 

What is AI? 

The British Veterinary Association (BVA) defines artificial intelligence as ‘the use of digital technology to create systems capable of performing tasks commonly thought to require human intelligence’. This definition comes from the accepted NHS AI Dictionary, as the terminology around what AI is and isn’t is confusing and inconsistent. 

The basis for many AI tools are ‘large language models’, computer programmes which are ‘trained’ on large datasets to repeatedly predict the next word in a sentence. The AI does not understand your question, it can just assess what words are most likely to go together and so give you an ‘answer’. This acts as the foundation model for ‘generative AI’ (like ChatGPT) which uses prompts to create original content including audio, images, software code, test or video, creating ‘answers’ that are similar, but not identical, to the original data it was trained on. 

What could AI be used for in veterinary medicine? 

AI has massive potential for improving standards in veterinary medicine. A few examples include: 

  • analysing x-rays and lab tests to provide more consistent and faster results 
  • bringing together all the information about an animal (history, lab results, imaging etc) to quickly produce overviews that can inform faster treatment decisions 
  • surveillance of emerging diseases 
  • production of client support materials that can be given out at the same time as a consultation 
  • increasing access to veterinary care through teleconsulting (if vet and client are not in the same room) and translation (if vet and client speak different languages) 

What are AI scribes? 

Veterinary AI ‘Scribes’ are purpose-built tools that listen to consultations, analyse them and then convert them to a standard format of notes for your animal’s clinical history. There are many products now on the market that are specific to veterinary medicine (a critical factor in safety) and have been trained only on peer-reviewed veterinary research and sources. 

AI scribes have actually been in use for a few years, and are now evolving into AI ‘co-pilots’. These will not only act as scribes but can create tasks (such as ordering blood tests), create visual guides (such as dental charts), send communications to other members of the veterinary team (such as writing referral letters to specialists) and create summaries and care plans for owners to take home. 

How might this benefit my pet? 

One of the greatest benefits of an AI scribe is that it allows your vet to concentrate on you and your animal during a consultation. Without worrying about scanning histories and writing notes, your vet can give you their undivided attention. 

Whilst vets will try to write up their notes as soon as possible after a consultation, our jobs are fast paced and with high demand, and notes are inevitably done by memory, anywhere from 5 minutes to 5 hours after the event. We are all human, and inevitably details will be lost or misremembered (much in the same way as when you get home from the vets, it’s hard to remember exactly what was said). An AI scribe removes this error by transcribing in real time so that every detail is recorded. 

Every vet will inevitably have their own style of writing up notes, and an AI scribe will use a more standardised format (typically ‘SOAP’ – subjective (the history), objective (the clinical exam), assessment and plan). This means more consistent treatment regardless of which vet you see at your practice, as everyone can clearly see what was done and what was planned as the next step. 

These programmes can also generate a summary of your appointment then and there, a written document about what your vet said, what they plan to do next and what you need to do at home. Our pets have many caregivers at home who are rarely all present for the consultation, so this can take the pressure off the person who was in the room, often being given lots of complicated information in a short space of time. 

What are the potential risks for my pet? 

AI is a tool, just like a blood machine, and the output it produces is meaningless if the person using the tool does not understand how it works or how to analyse those results. Dr Google and AI have not and will not replace vets for this very reason, as it is the understanding of what the results mean that is more important than the results themselves. We need to ensure that any vet using AI is well trained in how to do so safely, so they can analyse its output as they would do any other clinical test.  

Bias is a significant concern in all forms of AI, as these tools are still developed by humans who will always put their own training and beliefs into what they make. It is essential that any AI used in a veterinary setting was developed using trusted, peer reviewed, veterinary specific data (that is kept up to date).  

Vets also need to be very careful around use of client and patient data. If the information used for the AI is sent outside the practice (for example to external servers), data could be obtained legally or illegally for unknown purposes beyond what the client has consented to. Cybersecurity as part of increasing AI use is essential. 

Is anybody regulating how AI is used in practice? 

There is no legal framework for regulating AI in the veterinary world, unlike in human medicine. 

As a result, the British Veterinary Association (BVA) has produced a policy on the use of AI in the veterinary profession which has 8 key principles: 

  1. AI is a tool – Technologies should be used to support, not replace the vet.  
  1. AI literacy – Vets should understand the limits of AI and validate critically. 
  1. Vets in AI development – Vets must guide development and validation. 
  1. Awareness and management of bias – Identify data bias for ethical use. 
  1. Confidence and agility – Vets must stay up to speed on AI technologies and be confident in their use. 
  1. Consent, privacy and client trust – Ensure data privacy and client consent. 
  1. Human oversight and responsibility – Vets must oversee AI use and be responsible for decisions that are made. 
  1. Explainability – AI technologies must demonstrate how they reached their conclusions and the vet must understand this. 

There are also 8 key recommendations from their report, encouraging proactive engagement of all veterinary professionals with AI and how it works, with practice teams openly discussing it’s use and having well used safety policies. They recommend that developers should provide accessible information about how their product works, and that regulators both in the UK and abroad should actively regulate AI in practice. 

Clinical Use 

A key feature of the BVA policy is the acknowledgement that there is a sliding scale of how appropriate it is to use AI for different veterinary tasks, with some being low risk and some being unacceptably high risk. Administrative tasks and marketing carry very low risk, analysis of records and results carry higher risks, but any tool that aids clinical decision making carries a high level of risk and needs to be carefully monitored. Automated decision making made by the AI with no further assessment by a qualified veterinary surgeon is unacceptable. There is currently no regulation around this but is considered a key priority, and is likely to be a feature of the reform of the Veterinary Surgeons Act. 

Data protection 

Any tool made for veterinary use needs to be GDPR compliant, and clients should be made aware that a scribe is being used to record their consultation for the purposes of creating notes. 

Conclusion 

AI is a fact of the world we now live in, and has the potential to be a very useful tool in veterinary practice. In the small animal sector, it can help with making diagnosis more accurate and efficient, provide greater client care through undivided attention and the provision of summaries, and improve continuity of care through standardised and accurate histories. As with any tool, it must be understood and used carefully, and will never be a substitute for a vet’s knowledge and experience. Whilst regulation is still in its early stages, key professional groups such as the RCVS and BVA are actively exploring how the issue might impact the vets, clients and animals to ensure we use it as safely as possible. 

References 

BVA policy – Artificial Intelligence in the Veterinary Profession 

Generative AI vs. predictive AI: What’s the difference? | IBM