Using artificial intelligence (AI) in practice – advice for the profession
The use of artificial intelligence (AI) is becoming increasingly prevalent in both personal and professional aspects of life, and its development continues to advance at a rapid pace. These innovations present valuable opportunities to improve and expand access to veterinary care, as well as to support you as veterinary surgeons and registered veterinary nurses. There is a balance to be had, however, as overreliance or overconfidence in the abilities of AI risks the erosion of clinical skills and professional judgement.
It is important to use AI tools ethically, transparently, and appropriately to reduce the risk of unintended consequences which may have a negative impact on animal welfare and on public confidence in the veterinary professions. It is also important to acknowledge that use of these tools can present issues such as an inherent risk of bias, or hallucinations (i.e., confabulation or misleading information presented as fact).
The RCVS aims to support the veterinary profession in maximising the positive potential of AI, while also safeguarding against use in any manner that could potentially compromise animal health, welfare, or public health.
Whilst the Code of Professional Conduct (the ‘Code’) and supporting guidance do not specifically refer to the use of AI, we can apply the existing principles to the use of such tools. The advice below aims to summarise some of the key considerations to keep in mind when contemplating the use of AI in veterinary practice.
Incorporating AI into veterinary care
As veterinary surgeons and registered veterinary nurses, you remain professionally responsible for the decisions made when treating your patients – it would be inappropriate to wholly delegate clinical decision making to an AI tool.
AI can be used in a number of ways, including to free up clinicians’ time or to automate an administrative process, but there should always be a human in the loop taking responsibility for decisions. In practice this means that the human connection between you as the veterinary surgeon or veterinary nurse, the tool, and the output must be maintained.
AI tools can be used as an aid to decision making. However, in line with your Code obligation to provide appropriate and adequate care, you should scrutinise and critically consider any output in the same way as when you read a piece of clinical veterinary research or academic writing, e.g., is the source reliable? Is the dataset sound? Does the output make sense? What are the limitations? Are the assumptions accurate?
The Code states that you must ensure that you keep within your own area of competence and refer cases responsibly when further advice is needed. As such, the knowledge and experience of the person using the tool should be considered.In particular, whether the user has the requisite knowledge to enter the correct prompts to generate information and sufficient understanding of the subject matter to critically consider the value of the output.
The Code states that you must ensure that tasks are only delegated to those who have the appropriate competence. It therefore follows that tasks involving use of AI should only be delegated to those with appropriate knowledge and training. Lay colleagues should also be reminded to maintain client confidentiality and comply with practice policy regarding data protection when using the tools.
Data protection and client confidentiality
In the same way as other digital tools (e.g., the PMS) you should be mindful of the way the data, in particular clients’ personal data, is processed, managed and stored by the practice and by third parties. The Information Commissioner’s Office (the ICO) is the best place to seek advice on these points – you may find it useful to review the ICO’s guidance on AI. Being open and honest with clients about how AI is being used to process data in practice will be the simplest and safest way to prevent any compliance issues.
You also have a Code obligation to maintain confidentiality in respect of both client and animal data. You should not disclose client or animal data without client consent or another justifiable reason (see Chapter 14 of the supporting guidance) and as such, you should be mindful of how AI tools process and store data, including ensuring that the software developer does not have access to the data held by the practice and/or is not scraping or harvesting the data to train the AI tool without client consent.
Clinical and client records
Scribing tools are a popular AI technology used in both human and veterinary care. These tools can help save time by transcribing and summarising consultations. However, you have a Code obligation to ensure that you keep clear, accurate, and detailed clinical and client records, and you will be expected to ensure that any transcripts or summaries meet this standard, which means that AI generated records should be verified manually and any necessary edits made contemporaneously.
Chapter 13 of the supporting guidance provides an expectation around what should be included in the clinical and client record, and this guidance applies where scribing tools are used.
Be mindful that where AI has been used to aid differential diagnoses, clinical judgement should be used to ensure that only relevant information is recorded on the clinical record, as opposed to a generated list which may include conditions which have no bearing on the case.
Development of relevant knowledge and skills
The Code requires you to maintain and develop the knowledge and skills relevant to your professional practice and competence. Veterinary surgeons and registered veterinary nurses are therefore encouraged to familiarise themselves with the potential benefits and risks of AI. Before using AI tools in practice, it is crucial that you understand its limitations and when to challenge its output in order to ensure the safety and efficacy of the care that you provide to animals.
Implementing a new AI tool
Veterinary surgeons and registered veterinary nurses may benefit from considering the following non-exhaustive list of questions when implementing a new AI tool in practice or assessing the suitability of an existing tool:
- What is the scope of the software? How will the limitations affect veterinary care and how will this be communicated to staff in practice?
- How is the data stored? Does this comply with the practice’s privacy policy?
- How has the AI been trained? e.g., is the learning ongoing and will it involve live client data? If so, has the client consented to their data being processed in this way?
- Is the tool a closed system or is the data accessible to and editable by the software developer?
If you, or the software provider, are unable to answer the questions above, or if you have other concerns relating to the AI tool, you should consider whether using such software would be in line with your professional responsibilities under the Code.
Further information
If you have further questions about this guidance, please contact the Standards and Advice team via [email protected]