Just as there is a difference between riding in self-driving cars and using GPS navigation while still holding the wheel, artificial intelligence use in pharmacy spans a continuum, noted Ravi Patel, PharmD, the lead innovation advisor at the University of Pittsburgh School of Pharmacy.
A simple use might be autocompleting prescription information, saving time for providers. More complex uses of AI include support for medication management based on an individual’s clinical profile, to optimize drug dosages and minimize harmful drug–drug interactions.
“A few years ago, the questions in pharmacy centered around ‘what is AI?’ Today the focus has shifted to ‘how best can we use AI in practice?’” said Dr. Patel, who spoke at the APhA2025 meeting, in Nashville, Tenn.
One difference between self-driving cars and AI in pharmacy, Dr. Patel stressed, is that in pharmacy a human should always remain in the loop; AI is a decision aid, not a decision-maker.
Many AI tools rely on algorithms that predict a patient’s likely response to a drug and dose given data from similar patients. These algorithms need human oversight and tweaking, Dr. Patel said.
“What we call ‘algorithmovigilance’ is an emerging role for pharmacists,” he said (NPJ Digit Med 2024;7[1]:270). This is akin to pharmacovigilance, the discipline of mining real-world data to detect drug safety concerns that can be addressed with new guidelines or regulations. Algorithmovigilance involves adjusting AI algorithms to optimize the accuracy of their dosing or dispensing suggestions; Dr. Patel is working with other leaders in this nascent field.
While it may feel like AI appeared out of the blue with tools such as ChatGPT, Dr. Patel said clinicians have long worked with AI to some degree—for example, by using autocompletion of emails and of search phrases in internet search engines. Pharmacists rely on decision support tools, which are not different fundamentally from today’s AI-generated algorithms.
“We have long used clinical decision support tools to help guide dispensing and prescription practices,” Dr. Patel said, referring to software that gives pharmacists access to condition-specific drug order sets, evidence-based treatment guidelines, diagnostic tips and other actionable information. The difference today, with AI, is that far more clinical data can be mined for these tools than before. Such data mining leads to more customized dosing recommendations and decision support, sometimes down to which drugs are best for a given genotype or even what time of day is optimal for taking a medication.
This is not a bad thing, Dr. Patel said, as long as pharmacists—and not the computer—make the final call.
The sources reported no relevant financial disclosures.
This article is from the August 2025 print issue.