It is everywhere. Recently, I was talking to a gastroenterologist as he was heading off to work on a new AI assisted colonoscopy system, which helps to visualise and evaluate polyps. Machine learning has been with us for many years. Things changed dramatically when DeepMind started beating world chess champions. Last year a spin-off company, AlphaFold launched, that uses AI, and can predict protein folding with incredible accuracy. This is extraordinary technology, as it paves a clear path for the development of new variations of medicines and drugs.

Yes, it is quite obvious that, to date, AI has definitely not wheedled its way into community pharmacy in Ireland. Once it does, it would be fascinating to speculate how it may impact our daily working lives. For instance, how could AI make community pharmacy safer, more efficient? Currently, most of what we do has an enormous manual component to it. This, naturally, means that it is slightly more prone to human factors. This is why we have fairly rigid standard operating procedures to ensure both quality and safety. One area where AI could definitely make a huge difference is in prescription or medication review.

Currently, when we are presented with a prescription, our challenges are to determine, is it intelligible; is it legal; and, is it appropriate? For any AI system, it must be able to translate all the individual components into machine understandable language. So, if we start with the prescription, it needs to be in a form that the computer can readily understand. While many excellent optical character recognition systems exist, which would make a good stab at reading a prescription, I don’t think that we’re at the stage of trusting this yet. This, of course, is an issue that is going to be addressed as proper electronic prescriptions arrive. They will be structured and machine readable. Here in Ireland, we are already more than 20 years late to the electronic prescription party and, unfortunately, it’s not clear how soon the invites will finally go out. The legalities of a prescription should not be a difficult process, as there are specific rules about validity. Aspects like authentication, and technical issues like dates, are all readily addressable. Of course, this really means that the most important part of assessing a prescription, which is appropriateness and safety, is the Holy Grail. It is not difficult to see how any AI system could quite rapidly develop a safety model for community pharmacy. Indeed, many of the building blocks are already there. Yet, in the short to intermediate term, I would never see this as being anything other than a tool to improve processes.

Community pharmacy isn’t just about providing medicines to patients. It is much more than that. It is fundamentally about providing a health service to a public in need. As the world gets older, the pressures on pharmacy are going to increase. The last decade has demonstrated that the daily interactions, and complex work undertaken in community pharmacy, have increased massively in volume. In the last three years, the huge impact of increasingly systemic drug shortages has created enormous stresses. The complexity of the average patient interaction has increased substantially. Even with relatively trivial prescriptions, we are often faced with major challenges. Transitions of care have always been, and will continue to be, perhaps the biggest risk to public health from a community pharmacy perspective. Some of these threats have been around for a long time. For instance, the discharge process in a hospital is typically fraught with risk. Frequently, it is rushed, where the prescription is one of the final items delivered as the patient is going out the door. In most cases the community pharmacy will identify multiple queries on a discharge prescription. The prescribers are often uncontactable. Another growing threat is the simultaneous rise of the walk-in

clinic and the remote online prescriber. It would be fair to say that some of these prescriptions present challenges that require substantial resource is in the community pharmacy. I haven’t even begun to mention the unique set of challenges faced by Irish pharmacists in the form of the arcane State reimbursement system. Their growingly bizarre set of rules, grey areas, and head scratching processes are a constant source of annoyance, stress and bureaucracy for every pharmacist in the country.

Ironically, the greatest barrier to entry to AI in community pharmacy probably is a lack of quality data. We have been poor at systematically documenting the issues that we address. Large language models, the currency of AI, needs enormous amounts of data to give good results. So, while the computers won’t be taking our jobs any time soon, we should prepare for exciting developments.