The AI Engine Behind Smarter Care

Volume 14 June 30, 2025

✍️ From the Editor

As we head into the holiday week, we at Future Reps hope you've got some fun plans lined up! This issue takes a step back to better understand what’s going on inside the large language models (LLMs) that are quickly becoming part of our everyday clinical tools and how it effects you as a provider. As therapists, we know the value of building a strong foundation, and that’s our aim for this week. We're covering the cognitive effects of using ChatGPT, a plain-language explainer on reinforcement learning, and how Sword Health is using AI to improve PT scheduling at a time when more therapists are leaving the field than entering it. 

📈 This Week's Highlights

AI Optimizes PT Scheduling Based on Patient Risk Profiles

A recent analysis highlighted how AI is being used to reduce no-show rates and optimize PT scheduling. Using patient history, demographics, and behavioral data, predictive models can flag which patients are most likely to cancel or miss appointments. Clinics can then tailor scheduling or follow-up strategies accordingly.

🔗 Read more (3 min read)

Why this matters for PTs:
As the field is seeing a 3% regression annually having better scheduling practices to get the people who need it most in front of the people who can provide the best care is extremely helpful. Fewer no-shows mean better continuity of care and improved clinic efficiency. As more practices adopt these tools, PTs may gain new insights into how to structure follow-ups and maintain patient engagement.

Reinforcement Learning, Explained

This accessible explainer from Understanding AI breaks down reinforcement learning (RL), a foundational concept in many clinical AI tools. The article clarifies how RL agents "learn" by trial and error, optimizing actions based on rewards, discusses the future of RL, and draws connections to its use in real-world health scenarios.

🔗 Read more (4 min read)

Why this matters for PTs:
As we begin to use these tools having an understanding can help improve how the tools are utilized, as well as give you some firepower to ask the right questions of companies before implementing systems. If we’re using this for physical therapy, what physical therapy focused material has the system been trained on and does it aline with the users goals? From optimizing exoskeleton gait training to fine-tuning adaptive exercise devices, RL helps drive smarter, more responsive tools.

Your Brain on ChatGPT: Cognitive Costs of LLM Assistance

MIT Media Lab researchers conducted a four-month EEG-based study examining what happens when people write essays using ChatGPT, Google, or no tool. The study tracked brain activity across 69 participants and found that those using ChatGPT had significantly lower engagement in their brain regions tied to working memory, attention, and cognitive control. This was paired with a measurable drop in originality, memory recall, and a sense of ownership over the work, despite the writing appearing competent on the surface.

🔗 Read the study (5 min read)

Why this matters for PTs:
As AI writing tools creep into documentation, education, and patient communication, this study is a timely reminder to stay engaged. It suggests that while AI can save time, learning how to use it as a automation agent and not a “thinker” might stop the blunting of critical thinking and reduced information retention.

GPT-4o Personalizes HEPs for Different Learning Styles

OpenAI's GPT-4o can now respond in text, speech, and vision—all in real time. That opens new possibilities for customizing home exercise programs (HEPs) based on how patients best understand and retain information. For instance, GPT-4o could turn written HEPs into audio instructions, metaphor-based explanations, or visual demos.

🔗 Explore GPT-4o (3 min read)

Why this matters for PTs:
HEP adherence often hinges on how well patients understand their tasks. GPT-4o could help PTs create more tailored instructions without extra documentation time.

🔧 What’s New in ManagePT

Coming Soon: Integrated HEP Video Builder You’ll soon be able to build short video-based HEPs using our GPT-4o integration. Choose the exercise, select tone and learning style, and generate a ready-to-share clip.

Why it helps:
No more hunting for generic YouTube links. These videos can be personalized and embedded directly in your notes or patient messages.

📚 Book Recommendation

Ever wonder what Idaho potatoes and computer chips have in common?

Check out Chip War by Chris Miller. It’s not a beach read—dense at times—but it’s a compelling, well-researched look at how we got to this point in computing power and infrastructure. From Cold War geopolitics to modern supply chain battles, this book lays out why semiconductors are the backbone of everything from smartphones to smart prosthetics.

Why it’s worth your time: Understanding the history and fragility of chip technology helps contextualize why AI progress (and its limits) are so tightly tied to hardware. As healthcare becomes more AI-enabled, this backstory becomes surprisingly relevant.

Thanks for reading Future Reps. Got a tool, trend, or study we should cover next? Let us know!t trust.