Leaders from healthcare, patient advocacy, academia, informatics, and industry explore how AI could reshape patient safety, quality, and reliability. Moderated by Dr. Alan Forster, Vice President of Quality, Innovation and Performance at McGill University Health Center, this panel brought together Dr. Sean Miller, Dr. Joe Smith, Susan Sheridan, Dr. David Bates, and Dr. Namita Mohta for a timely conversation on one of healthcare’s most fundamental responsibilities: keeping patients safe.
Their message was both optimistic and practical: AI will not eliminate harm on its own. But if implemented thoughtfully, it can help patients, clinicians, and health systems detect risk earlier, close safety gaps faster, and make care more reliable.
Patients Are Already Using AI
Susan Sheridan opened with a powerful patient perspective. For her, AI is not something that will only be used by healthcare systems “on” patients. Patients themselves are already using it, often out of necessity.
She shared a personal story of experiencing facial drooping, pain, and numbness, only to be dismissed after an emergency department visit. On the way home, she and her husband turned to AI, which raised the possibility of Bell’s palsy and highlighted the urgency of treatment within a narrow time window. The next day, she received the appropriate diagnosis and treatment.
For Sheridan, this experience reflected a larger shift. AI is democratizing access to medical knowledge and helping patients become more informed, more prepared, and more active partners in their care.
Her call to healthcare leaders was clear: do not discourage informed patients. Instead, provide safe tools, help patients understand how to use them, and involve them in the development, oversight, and governance of AI systems.
From Digital Burden to Clinical Copilot
Dr. Sean Miller reflected on the early promise of electronic health records. Twenty-five years ago, digital records were expected to improve safety by reducing handwriting errors, improving documentation, and supporting clinical decision-making. While EHRs brought important gains, they also created new burdens and often made it harder for clinicians to connect with patients.
AI, Miller argued, feels different.
Clinicians are more engaged, more informed, and more eager to shape the tools being built. Ambient AI scribes, for example, are not only a burnout solution. By allowing physicians to listen more fully and capture better histories, they may also support safer diagnosis and better patient care.
Miller also highlighted physician-led AI applications already being used to close care gaps, such as identifying pregnant patients at high risk for preeclampsia who may benefit from aspirin, and AI-enabled radiology workflows that alert care teams more quickly in urgent cases such as stroke or pulmonary embolism.
His central point: AI is most powerful when clinicians are at the table, defining the problems and guiding the solutions.
Detecting Harm Before It Happens
Dr. David Bates brought the conversation back to the scale of the safety challenge. Patients admitted to hospitals still face a significant risk of harm, and complex systems create risks that are often difficult to detect in real time.
Bates described three practical areas where AI can already improve safety.
The first is clinical decompensation. AI can analyze large volumes of patient data and identify subtle signs that a patient may be worsening before clinicians would otherwise notice.
The second is follow-up on abnormal radiology findings. AI can scan radiology reports, identify concerning findings, and trigger processes to ensure they are reviewed, acted upon, and not lost in the system.
The third is real-time measurement of safety using data from the electronic health record. Instead of waiting for a serious event to happen and then reacting, health systems can use AI to identify patterns, detect risks, and prioritize the most common and preventable sources of harm.
In Bates’ framing, AI can help healthcare move from retrospective safety management to real-time safety intelligence.
Solving the Last-Mile Problem After Discharge
Dr. Namita Mohta focused on one of the most vulnerable moments in the patient journey: the transition from hospital to home.
As a hospitalist, she described how much effort goes into stabilizing patients, adjusting medications, completing tests, and preparing them for discharge. Yet once patients leave the hospital, care can quickly become fragmented. Instructions may be unclear. Medication changes may be confusing. Home risks may go unnoticed. Responsibility between the hospital team and the primary care team may become ambiguous.
Mohta described this as a “last mile” problem.
AI could help bridge that gap by translating discharge instructions into the patient’s language, adapting information into formats patients can actually understand, improving medication reconciliation, predicting risks in the home, and creating clear narrative summaries for the entire care team.
The goal is not simply better documentation. It is continuity: helping patients and care teams carry the plan forward safely after discharge.
Democratizing Expertise and Standardizing Care
Dr. Joe Smith offered an industry perspective, grounded in the reality that no clinician can keep up with the full complexity of modern medicine alone.
He described AI’s potential to help clinicians answer a critical question at the bedside: what is the next best step for this patient? Instead of relying only on one clinician’s training or local experience, AI could draw on broader datasets and global patterns of care.
Smith framed this as a democratization of expertise. AI-enabled tools could help clinicians recognize rare conditions, accelerate training in image-based specialties, and make expert-level pattern recognition more widely available.
He also pointed to the opportunity to standardize bedside procedures that remain highly variable today, such as IV insertion, dressing changes, and medication administration. With AI-enabled computer vision, health systems could potentially monitor whether procedures are being performed correctly, provide feedback, and support continuous education.
For Smith, this kind of always-on safety support could help move healthcare closer to true harm reduction across every room, every patient, and every day.
Start Now, Evaluate Quickly, Stop What Does Not Work
When asked what advice they would give healthcare leaders, the panelists emphasized discipline.
Miller urged leaders to take a problem-oriented view. Instead of being overwhelmed by distant future scenarios, organizations should focus on urgent safety problems that need to be solved now, then identify AI tools that can help address them.
Bates encouraged leaders to choose a few interventions, test them, and spread what works. Just as importantly, organizations need the discipline to stop doing things that are not producing results.
Smith added that AI is not a lake you can wait to jump into later. It is a fast-moving river. Healthcare organizations need to begin building the internal capability to evaluate tools rapidly, decide what is good enough to keep, and replace what is not.
Miller also emphasized trust. Patients and clinicians will need confidence that the AI tools deployed by healthcare organizations have been properly evaluated, governed, and monitored. Trust in AI will not be automatic. It must be earned through transparent processes and responsible implementation.
The Takeaway
The panel closed on a note of cautious optimism. AI is already being used by patients, clinicians, researchers, startups, and health systems. It is moving quickly, and in some areas healthcare may be adopting it faster than expected. But speed alone is not enough.
To improve patient safety, AI must be guided by clear priorities, strong governance, patient involvement, clinician leadership, and rapid-cycle evaluation. It must help healthcare systems detect risks earlier, act faster, reduce variation, and support safer transitions across the care journey.
Most importantly, AI must strengthen, not replace, the relationships at the heart of safe care. The future of patient safety will not be defined only by better algorithms. It will be defined by how well healthcare leaders use those tools to empower patients, support clinicians, and build systems where harm is identified, prevented, and learned from before it becomes invisible.
