AI in Pharmacy Applications – beware the rise of the machine!
I should start by saying that I am a fan of AI.
I remember having to collate a lot of PDF documents, with lots of numbers on them, into an excel spreadsheet. I put it off as long as I could and the allocated half a day to the very boring task. Then I had a thought, ChatGPT! I uploaded the documents, told it what to do, and 5 minutes later my spreadsheet was ready. I even checked the numbers and they were perfect. Half a day saved and I was a convert.
Unfortunately in other areas things have not gone so well. Travelling through Barcelona airport I wanted to know if they were using the new scanners and AI told me that they definitely were, which was great. Unfortunately the terminal did not have a single new scanner and was still using the 100 ml rule as a result, so I had to throw away some expensive liquids and repack everything after finding a small plastic bag. AI identified my coffee machine from a photo I took and told me there was no way to adjust the water temperature flow. Weeks later I found the manual and it explained exactly how to adjust the temperature.
So I am a fan, but it’s all “subject to verification” now.
Which brings me to pharmacy applications. Even a year or so ago I had not seen AI being used by Applicants or Objectors, but now things are different and it is a really mixed bag or results. I recently had a client send me an email with lots of suggestions for how I could make his submission better. It was obvious in a few seconds that the response was AI generated. I have no problem at all with that, but the response stated with absolute confidence that there were certain things that I needed to do. Unfortunately every one of them was wrong, and in some cases, total nonsense.
In another case I was asked to help a pharmacy contractor object to an application to open a new pharmacy fairly close to their existing pharmacy. The Applicant had engaged their own consultant to help with the application and no doubt paid them handsomely for their work. The application looked very professional and read very well. The Applicant had set out the legal test and referenced a Court of Appeal decision to support their interpretation of the law. There was no room for doubt and the submissions were clear. Unfortunately the submissions were complete nonsense and did not correctly reflect the law. I happen to have been involved in the Court of Appeal decision that was referenced and knew immediately that what was being stated with total confidence was also total nonsense. My response highlighted that the arguments being used by the Applicant appeared to be AI generated and the ICB refused the application. The Applicant chose not to appeal, which was probably wise.
In a recent Distance Selling Pharmacy application in which I was acting for the Applicant, we had a lot of objectors. All but one used AI to generate their objections. The objections were absolutely clear that there were several reasons why the application had to be refused and every single one of those reasons was nonsense.
In another case I was acting for an Applicant who had received ICB approval for their application and was now at appeal stage. I wasn’t involved in the initial stage but they wanted help with the appeal. My client sent me his proposed response to the appeal and it was obviously AI generated. The AI tool had listed all the reasons why it thought the appeal points were deficient or irrelevant, but totally failed to address the reasons why the application should be approved at appeal – which was the most important thing to do.
So I have some words of caution. If you feed this type of work into an AI large language model and ask it to help you perform a certain task, then it will. The suggested response will be very well put together and will look impressive. If you do not understand the Regulations properly then you will be blown away by just how good the results are and entirely unaware that a lot of it may be nonsense.
AI can be really helpful. I use it. I am quite sure it will replace a lot of human work, especially in the consulting and legal sectors but probably everywhere. Change is happening, but now people are using tools without the ability to verify if the responses are relevant or correct and there is a lot of blind faith.
And I know people will say, “well he would say that, wouldn’t he!”, and that’s fine.
Conor Daly
Partner at Rushport Advisory LLP