Artificial intelligence (AI) technologies are increasingly being integrated into clinical workflows, with pharmacists implementing solutions to transform patient care, optimize medication use, and streamline operations. ASHP launched an AI case study library to showcase real-world examples of how members are using the technology creatively, effectively, and responsibly — and to inspire others looking to incorporate AI into their pharmacy practice.
To gain even deeper insights, ASHP spoke with three case study authors:
- Taylor Beauchamp, PharmD, (former) PGY2 ambulatory care pharmacy resident with Legacy Health, Portland, Oregon, who reviewed the adoption of an AI-assisted, automated clinical documentation tool within the pharmacy team
- Robbie Freeman, DNP, RN, chief digital transformation officer, Mount Sinai Health System in New York, whose team developed and deployed an AI-powered risk stratification tool targeting older adults in the ICU, to address the frequent underdiagnosis and delayed recognition of hospital delirium
- Collin Lee, PharmD, BCPS, FGSHP, corporate director for clinical pharmacy services, Emory Healthcare, Atlanta, Georgia, who piloted an online pharmacist-reinforced AI tool to help answer complex, time-intensive drug information questions and formulary reviews.
Read on to learn how they keep up with emerging technology, how they made the business case for AI, and what surprised them along the way.
Note: Responses have been edited for length and clarity.
How has this use of AI impacted your pharmacy practice?
Beauchamp: We’ve seen measurable improvements in key areas like provider burnout, time spent taking notes, and the number of appointments per day. The AI tool gives us extra time to spend with patients or on other work.
Freeman: Outcomes have been substantial: delirium detection rates increased more than fourfold, and daily doses of high-risk deliriogenic medications have been significantly reduced. My colleagues and I detailed our methodology and results in a May 2025 JAMA publication. Dr. Joseph Friedman is the clinical brains behind the vision for using AI for delirium prevention and management, and my team is the technical brains that helped bring that idea to life.
Lee: Our AI-driven tool saves us a ton of time. Right now, I might do a regular search on a medicine that has been around for a long time and find 400 randomized controlled trials — it’s overwhelming! The tool runs a drug information query through an algorithm using the literature. Then, pharmacists read that answer; their feedback helps build a bigger and better AI tool. We like knowing a pharmacist is looking at [the responses] and it’s not just AI coming back to us.
How did you get buy-in from your health-system leadership for this project?
Beauchamp: SBAR [Situation, Background, Assessment, Recommendation] became our go-to framework. To obtain the AI tool license, our clinical pharmacy manager, Ryan Wargo, had to clearly outline [pharmacists’] impact, our patient volumes, and our support to providers. To obtain additional licenses, we used our research data and highlighted the positive effect on burnout, which is where Legacy providers have seen the biggest impact from the AI tool. It’s important to track metrics in language that leadership speaks and understands.
Lee: It helped that the tool that we’re using is not completely AI. I’m not sure that we’re at the point yet that we want to fully trust AI with our medical information. The company also helped me with the proposal because they keep data on usage and time savings, and how that translates into cost savings.
What surprised you the most about your AI project?
Beauchamp: I was surprised by two things: how accurately the AI tool interpreted patient conversations and how easily our pharmacists were able to integrate it into their practice. Our pharmacists felt that the AI-assisted tool did a great job of excluding non-health-related conversation while still keeping an accurate record of the clinical visit. And our pharmacists also adapted to it pretty quickly. At the one-month mark, about 20% of our pharmacists’ clinical note content was being generated by the tool.
Freeman: We got so much value from the clinical notes. Ours is what we call a multimodal AI model, meaning it uses different types of data streams. The initial version used structured data like laboratory values, vital signs, or flow sheets. When we added unstructured free text, it created a much more powerful tool. Generally, the more data, the better when it comes to AI tools.
Lee: I was surprised by how much pharmacists used the AI tool. They’re asking direct, specific questions about drug stability, drug interactions, vaccine indications, alternatives for drug shortages, all kinds of things. Not only that, but they’re searching the library of questions and answers from health systems around the country to keep up with what’s going on in the world [of pharmacy].
What’s something else you learned?
Beauchamp: I learned phrasing has a lot to do with [acceptance] because AI is a pretty controversial term now. We ask for consent at the beginning of every patient visit because patients are being recorded with the AI-assisted tool. We made sure to explain why we wanted to use it: so we can focus on you during this visit. Patients were generally amenable to that because that’s something that they really want from their providers.
Freeman: I’m always humbled by how much we need to focus on the workflow if we want people to work differently. We really have to get into the nitty-gritty: How do people want to see it in the electronic medical record? What are our suppression criteria? We also want to explain the ‘why.’ It always comes down to people, process, and technology — and people and process are the hardest parts.
What’s the No. 1 thing you want your pharmacy peers to take away from your case study?
Beauchamp: The only way we’re going to push the pharmacy profession forward is by advocating for ourselves and ensuring that we have the same tools to serve our patients as other provider types. It’s important for us to do that now so that the generation of pharmacists behind us doesn’t have to. It starts with little things like [this AI-assisted notetaking tool].
Freeman: AI can help improve safety and outcomes, but don’t underestimate the ability to reduce workload burden. If you could start over and redesign a process with this new set of tools and capabilities, you probably wouldn’t design it exactly as it works today. I don’t think AI is going to replace clinicians, but it certainly could give us new superpowers that we didn’t have before. It’s good to have some fundamental ability to read and understand the code, but you don’t necessarily need to have an advanced PhD in computer science. If you have a hunch, you could use one of the frontier AI tools to help build models that work pretty well, especially for exploratory analysis. Just make sure you have the right guardrails and agreements in place before putting in protected health information.
Lee: I think we’re to the point that AI is an aid, but not to the point where it’s going to take over. We can’t just accept AI without having some type of medical eye on it, making sure that the answers that we’re getting are safe to take action on.
Did this experience dispel any misconceptions you had about AI?
Beauchamp: When I first started using the AI tool, there was this idea that it was going to replace how pharmacists wrote their notes. But we still see that manual typing is necessary for editing and formatting; AI is not intelligent enough to notice your style and conform to it. The AI tool generates text in paragraph form, while we like to use bullet points that are very concise and easy for other providers to read. If you use an AI tool like this, there is going to be a compromise between that and efficiency.
Lee: I learned that the AI tool was better at some tasks than others. It was very helpful in quickly answering drug information questions. It was less helpful drafting monographs for our pharmacy and therapeutics (P&T) committee; we had to do a lot of fine-tuning for Emory-specific data. So it didn’t save us as much time on the monograph side as I thought it would. We still have to be the experts on that drug.
What’s next for AI in your practice?
Beauchamp: We got five additional AI tool licenses and started collecting data on those. When the six-month trial runs out, we’ll be able to talk to executive leadership again about expanding the number of pharmacists who have access to this AI tool.
Freeman: Mount Sinai has been building our clinical data science team for about seven years, and the delirium model is part of a larger portfolio of AI products. We’ve been maturing our ability to go from an idea to scoping it, building the model, and rigorously testing it in a real-time environment. Generally, it takes us about a year to go from an initial idea to a live product that’s being used at scale. We’re always trying to see how we can improve there.
Lee: We are still piloting this AI tool and trying to decide whether we want to bring it in full-time. We’re weighing the cost of the tool versus the cost of having another pharmacy employee, who you can utilize and reallocate for anything. We’re also testing out another generative AI tool that can conduct observational studies of real-world health databases to give us some benchmarks for decision-making.
There is a great deal to learn about the role of AI in healthcare. How do you keep up with it all?
Beauchamp: I’m pretty tuned into technology and use AI frequently in my personal life. I actually found out about this AI tool during my residency orientation, and that’s what spurred the idea for this research project. It helps to keep up with what’s going on in your own health organization, because most likely, they’ve got some AI project going on in the background. I also keep up with AI-related content posted by ASHP and other pharmacy organizations.
Freeman: We learn a lot through doing, and each project we learn something new. The technology is certainly impressive, but a lot of the value happens on the human stack —meaning, working with people on the other end of the AI tool to make sure that it’s going to fit in seamlessly. We have this co-design process that includes the voices of the people using the tools. Just like you would never build an airplane without feedback from the pilots, we can’t build healthcare AI without feedback from our clinicians. One helpful resource I recommend for pharmacists interested in AI is the Coalition on Healthcare AI.
Lee: Present on it! My drug information pharmacist and I gave a presentation at the Georgia Society of Health-System Pharmacists on how AI can enhance evidence-based medical decision-making. We forced ourselves to learn about AI, and now I feel like I’m invested in it.
ASHP is highlighting case studies to show how AI can enhance patient care, optimize workflow, and drive innovation in practice and education. View ASHP’s growing case study library and consider sharing your own AI story to help your colleagues learn from your experience.
