Preparing Pharmacy Students to Lead in the Age of AI

As the pharmacy profession adapts to new technologies, educators face the challenge of preparing students in a rapidly changing landscape.   

Conan MacDougall, PharmD, MAS, an infectious diseases pharmacist and professor in the Department of Clinical Pharmacy at the UCSF School of Pharmacy, has studied how AI tools perform in clinical problem-solving. 

Tomorrow’s pharmacy leaders, MacDougall says, will be defined not by the tools they use, but by how thoughtfully they use them — balancing technical expertise with communication, judgment, and patient-centered care.  

Q: You recently co-authored a study testing ChatGPT on infectious disease questions. What did you find?  

A: We created a set of 100 questions that represented the kind of difficult, cutting-edge questions we answer as infectious diseases pharmacists. Then we ran them through the best version of ChatGPT available at the time and had experts grade the responses. About 40 percent of the time, the answers were useful. Other times, on what I thought were simpler questions, it gave bad — or even harmful — answers.  

But given rapid advances in AI, by the time our study was published, it was already out of date. Other researchers repeated it with a newer version, and their results showed about 90 percent usefulness.  But even if 90 percent are right, that still means 10 percent are wrong. And the question becomes: Who decides which 10 percent? That’s where you always need people with high expertise to verify the answers.  

Q: How does that change the way you think about teaching and assessing pharmacy students?  

A: As an educator, you can no longer be sure that the work students turn in wasn’t generated by AI. The tools we have to detect AI writing don’t work very well, so I think we have to reorient.  

We’re going to need more real-time discussion, more Socratic questioning, and more back-and-forth with students to understand what they actually know. That’s generally a good thing, but it’s harder to scale. Education is going to have to adapt. 

Q: Looking ahead, what leadership capabilities will pharmacy students need beyond drug knowledge?  

A: There will always be a need for people with some basic knowledge of AI but with deep content expertise to check and validate what AI produces. A new path will grow for people with basic clinical knowledge but deep AI expertise who can work with these systems, deploying them effectively and knowing where they fit in practice.  

No matter which path you’re on, the human skills matter most. These models don’t really understand context. They can say, ‘This is the best drug according to the guidelines,’ but they can’t weigh a patient’s living situation, or their preferences, or how a treatment fits into their overall care. Communicating, contextualizing, and working as part of an interdisciplinary team are the leadership skills pharmacy students will need.  

Q: You also studied how students best learn the fundamentals of “bugs and drugs.” How do you see AI playing a role in antimicrobial stewardship, which is one area where pharmacists already lead?  

A: At UCSF and elsewhere, people are studying ways to train AI systems to think like infectious diseases pharmacists. The way it usually works is as a human-in-the-loop system. The AI scans all the antibiotic use in a hospital and flags patients who might benefit from a review. Then the pharmacist looks at those cases and decides if an intervention is warranted.  

That’s useful, because at any given time about half of hospital patients are on antimicrobials. And antibiotics are unique. Using one patient’s antibiotic can affect another patient through resistance. Stewardship means balancing what’s best for the patient in front of you with what preserves these drugs for the future. AI may help us spot opportunities, but the pharmacist’s judgment is still essential.  

Q: You’re developing AI “tutor bots” for pharmacy education. How will they help students learn? 

A: Probably the single most effective educational intervention is one-on-one tutoring. But we can’t give every student a personal tutor. AI gives us a chance to get closer to that.  

In a course I teach, I built a custom GPT to act like a tutor, not just a summarizer. Instead of saying, ‘Here are the five key points on this topic,’ it asks, ‘What do you already know about this topic? Let me quiz you. Let’s work through this together.’ More than half the students used it, and 80 percent of those said it was valuable for their learning.  

Now we’re taking that idea further. With funding from UCSF, we’re developing tutor bots on the university’s secure Versa Chat platform, which is free and unlimited for students and is compliant with student privacy regulations. We’re also creating a kind of cookbook so other faculty can design tutor bots for their courses. The goal isn’t to replace faculty. It’s to give students an additional resource that helps them practice, get feedback, and build the mental models they need. At the end of the day, the key to learning is practice with feedback. AI gives us a way to scale that in a way we couldn’t before.