In his role as Trinity Health's executive vice president and chief operating officer, Dr. Dan Roth frequently leads presentations on artificial intelligence integration in healthcare. They often use a similar descriptive phrase in their titles: Promise and Peril.
"It's our job to maximize the promise and minimize the peril" in the adoption of AI tools, he explained.
It's a goal shared by many Catholic health system leaders across the country who are working to balance the vast possibilities of an AI-driven healthcare transformation with a duty to responsibly adopt emerging AI tools.
"We know that this is an opportunity for us to be transformative in how we provide care," said Roth. "But we don't want to do technology for technology's sake. We need to make sure we're focused on the areas that are the highest 'pain points' for our patients and our caregivers." The Livonia, Michigan-based system he works for provides services in 23 states.
A LISTENING TOOL
Catholic health systems across the country now commonly use ambient listening products with embedded AI to automatically generate patient visit notes, particularly in their primary care and other outpatient settings. Products such as Microsoft's DAX Copilot, which can integrate with popular electronic health record (EHR) platforms, like Epic, are reshaping the patient-provider connection in positive ways.
With an ambient listening scribe activated, physicians no longer have to talk to patients while typing notes at a computer. They can get back to having true face-to-face dialogue, focused on building a "healing connection of understanding about what might be going on," said Byron Yount, PhD, chief data and AI officer and vice president of transformation operations at Mercy, which serves patients in Arkansas, Illinois, Kansas, Missouri and Oklahoma.
At Trinity Health, roughly 1,500 outpatient providers are currently using ambient scribing. "We are using it at scale, and we're looking to expand it to an inpatient setting for our hospitalists and nurses and even into the emergency room," Roth said. "Its scope will continue to grow."
Patients are noticing the difference — and say they appreciate not having to talk to their provider across a laptop screen. With ambient listening in place, patients reported "they felt for the first time [in recent memory] that their providers were fully engaged in listening," said Dr. Ann Cappellari, chief medical information officer at SSM Health, headquartered in St. Louis. "It's a tool that can help restore that sacred provider-patient construct. And fostering that relationship is huge for healing."
Systems also report success using AI-supported tools to efficiently create patient treatment notes and discharge summaries across their inpatient settings. Not just a time-saver for clinicians, the resulting documents also serve as a helpful resource for patients who can refer to them as needed in their online patient portal to stay informed about each step of their care.
"It helps everyone all around," Cappellari said. "The AI-supported summaries include citations that have rigor and provenance. Patients can refer to their summaries and see the labs that they got, any prescription changes and their next care directions."
Health executives predict that, in time, AI-driven ambient listening tools and care note summarization tools will become even more integrated and robust. In the future, they will be able to lay out and summarize follow-up reminders for clinicians and provide even more helpful contextual data for patients.
When a patient gets a lab result back in their patient portal, these reports may soon offer additional interpretive context about what their cholesterol or other test findings may mean, Roth explained. "This is an AI-supported function that's coming in the not-so-distant future — and it's one that can help patients better navigate their own care journeys."
NEW FRONTIERS OF ADOPTION
Systems like Trinity Health, Mercy and SSM Health have also successfully deployed AI-supported tools to augment their radiology and imaging services.
"Radiologists are still reading every study, but they're doing it in a way that's faster and better because of the technology," Roth said. "The AI is empowering and enhancing the work of our professionals."
Given AI's dexterity at pattern recognition, it can excel at spotting and flagging even small abnormalities on an X-ray, MRI or CT scan and identifying which images should receive immediate review. "It enhances the discovery of nuanced things. It can prevent tired [human] eyes from potentially missing something," added Mercy's Yount.
Clinicians say AI, when well-integrated into healthcare, will do more than increase ease of record-keeping and speed up diagnostics. They're seeing ways it can aid in identifying illness and streamlining treatment when time is of the essence.
As a result, many systems are exploring how AI-supported services might offer a pathway toward earlier cancer detection. Health executives report early forays into using AI systems to review mammography images, as well as lung and colorectal cancer screenings, among others.1
"Lung cancer screening involves rapid, low-dose CT scans that identify small pulmonary nodules, many of which will need follow-up," explained Dr. Thomas McGinn, senior executive vice president and chief physician executive officer at CommonSpirit Health, headquartered in Chicago. "AI tools can immediately recognize which nodules are at a lower and higher risk [for cancer development], and whether they're changing from six months to a year between scans."
Trinity Health, CommonSpirit, SSM Health and others have also found success utilizing AI to speed stroke detection.2 AI-backed platforms, including RapidAI, Brainomix and Viz.ai,3 can help clinicians more swiftly assess patients who present with stroke symptoms — and offer data insights to inform best pathways of care — potentially lowering patients' risk of long-term stroke-related impairment or death.
"We've seen tremendous benefits from our acute stroke [AI adoption] programs," McGinn said. With appropriate AI-backed tools in place at community hospitals, "imaging is immediately transferred to the stroke center, so that stroke teams can be activated and a treatment plan developed more quickly than ever before."
Cappellari agreed that AI's capacity for stroke detection has been transformative. "It can detect a stroke within seconds of the CT getting done," she said. "It then pushes a ping to the radiologist and to an interventionalist," allowing the care team to quickly facilitate appropriate treatment.
At Mercy, an AI tool called Sepsis ImmunoScore4 will support providers in identifying and flagging hospitalized patients who may be at higher risk of developing sepsis. Several systems also use AI-supported tools to help clinicians evaluate when patients may be ready for hospital discharge or to identify patients with a higher-than-average likelihood of hospital readmission due to chronic illness.5
Meanwhile, Roth points to AI-EKG tools as another area ripe for expansion. "We are discovering that [with support from AI] we can get information from EKGs that we never thought was possible," he said. New research shows that AI tools can analyze EKG results to detect patients at high risk for heart failure and other heart diseases.6
MOVING FORWARD FOR PERSONALIZED MEDICINE
Many of the AI uses in healthcare today focus on "being more proactive and personalized and predictive with care," said Yount. He believes AI will increasingly improve its ability to both support earlier disease detection and identify optimized prevention plans and treatment options for patients, based on their genetics, disease risk factors and other key health metrics.
"We are seeing that AI can help us understand what types of treatments might be more effective and applicable for a given individual, and that type of application of AI is only going to get better over time," Yount said.
Yount was involved in helping create an in-house predictive AI tool in use at Mercy that evaluates electronic medical records to identify and flag patients at increased risk for various diseases, such as colorectal cancer. Once the system identifies these patients, their providers can then engage them in discussions about the importance of cancer screening and the right level of screening given their risk.
"The [AI] algorithms help you target your diagnostic tools more efficiently and effectively," Yount said.
AI's ability to manage large troves of data makes it an especially helpful tool as hospitals navigate increasingly complex cancer screening guidelines, CommonSpirit's McGinn said. He shared, as an example, the ways that recommended screening guidelines for various cancers can differ by age, depending on a patient's health history and prevalence of cancer in their family. Multiply that across all cancer types and among thousands of patients, and it can feel difficult for human providers to keep up.
To address this challenge, CommonSpirit uses robotic process automation, a software solution complementary to AI, to automatically calculate patients' personalized risk for key diseases based on their EHR data. Right now, the system is focusing on breast, colon and lung cancers, and the process is programmed to "automatically offer a screening test to be approved by the physician" for patients deemed at high risk, McGinn said. "We've seen huge increases with our cancer screening as a result, in some areas as much as a 50% increase year to date."
At Providence, teams created their own in-house AI tool, called Provaria, to facilitate more seamless patient messaging management.7 The system categorizes and directs patient messages in MyChart to appropriate response teams, allowing doctors to quickly identify and respond to patients who need immediate care first, while directing less urgent queries — like scheduling issues — to front office support staff.
The tool "reads" incoming patient messages using AI, then assigns them a triage category accordingly, explained Dr. Ford Parsons, associate vice president of clinical informatics and chief medical information officer of AI and engineering at Providence, headquartered in Renton, Washington.
The result: Patients whose symptoms warrant an appointment often see a provider the same day they send their message.
"With symptom messaging, our turnaround time [using Provaria] decreased by almost half," Parsons said. "So, from a patient perspective, they're getting information back much more quickly."
DATA-DRIVEN CARE ADVANCES
Multiple Catholic healthcare systems — including Providence, Trinity Health, CommonSpirit Health and Bon Secours Mercy Health — have partnered with Truveta, an AI-powered data intelligence company launched in 2021, to develop a comprehensive database of more than 130 million patients' de-identified EHR data.8
The Truveta platform gives its 30 partnering healthcare systems access to research-ready, anonymized patient data, which can be leveraged to assess and improve patient care quality, analyze the effectiveness of emerging therapies or track longitudinal patient outcomes, among other applications.
Truveta was "founded with the notion that we can use AI and computational systems to solve [healthcare] challenges through deep, rigorous research using de-identified patient data," said Dr. Michael Simonov, the company's senior vice president of product.
Truveta's newly launched Truveta Genome Project aims to create the "largest and most diverse de-identified database of genotypic and phenotypic information ever assembled," according to the company.
The project's large-scale scope will allow it to create a dataset that's far more representative and inclusive of diverse ethnic and sociogeographic groups than any before assembled — leading to more equitable treatments and care predictions, the company's leadership believes.
"It's an unparalleled investment to sequence millions of genomes for patients," Simonov said. "It's a future-looking perspective to the world of precision medicine, where we will treat every patient as distinct."
With future insights from the Truveta Genome Project, clinicians may eventually guide each patient's unique clinical care — and even predict or prevent certain disease progression — on the basis of their genetic makeup.
At Mercy, the system has pursued global partnerships in data and AI innovation, Yount said, including its founding partnership with Mayo Clinic Platform. This collaboration combines clinical experience, technical capabilities and decades of de-identified patient data from its partner organizations.
By drawing on large and diverse populations, this work focuses on identifying risks earlier, forging more representative studies, and developing new diagnostics and medical interventions in ways that no one system can achieve alone.
Mercy clinicians are already using AI-driven insights from this partnership to better understand "how different phenotypes can be treated more effectively with different types of medicine and medicinal interventions," Yount said.
As AI capabilities progress, "it will be able to predict for physicians, 'These are the patients whose diabetes will likely progress,' or 'These are the patients who will be more likely to have kidney problems or eye problems,'" Roth said. "We believe that [with AI] we will be able to learn and be better predictors of which patients need more focused support and interventions."
ETHICS AND MONITORING
Even with all of AI's promising potential — or, rather, perhaps because of it — health system leaders are treading cautiously and deliberately as they work to advance AI usage across their footprints.
At CommonSpirit, the health system has assembled a robust group of ethicists, administrators, clinicians and AI specialists that meets at least weekly to review performance metrics from its current AI usage and to evaluate the merits of new potential applications.
"It's a way of saying, 'We're embracing this, but we're embracing it in a thoughtful way,'" McGinn said.
During these ethics, data, algorithm and governance reviews, potential new AI tools undergo a thorough assessment of their diagnostic, therapeutic and preventive capabilities. That's particularly true for any applications that are close to bedside or that engage with patients in any way, McGinn said.
Providence, similarly, engages AI review council workgroups composed of doctors, data scientists, informaticists, nurses and ethicists who, together, evaluate workflow and adoption of AI to "make sure that we're getting the benefits of AI without any of the harm or safety risks," Parsons said.
When AI adoption is done with intention and within frameworks of ethical use, it has the power to provide patients with "a straighter line to care," said Yount. He added that at Mercy, a key goal is to ensure AI is adopted equitably, so that both the most resourced and the least resourced can benefit from its potential.
Cappellari also noted excitement about AI's potential to transform healthcare for the better. "It actually feels like [AI] is a care changer," she said. "It seems to have mass potential for benefit, patientwise, clinicianwise and health systemwise."
Even as AI advances, Cappellari notes that it will never take the human out of the loop. Indeed, most executives agree that when done well, AI has the power to let humans excel at what they do best: taking care of one another.
"If we use AI thoughtfully with clinician involvement, we can actually enhance the human connection of medicine," Parsons said.
ROBIN ROENKER is a freelance writer based in Lexington, Kentucky. She is an experienced reporter who often writes on health and wellness, higher education and business trends.
NOTES
- Omofolarin Debellotte et al., "Artificial Intelligence and Early Detection of Breast, Lung, and Colon Cancer: A Narrative Review," Cureus 17, no. 2 (2025): https://doi.org/10.7759/cureus.79199.
- Dr. JoonNyung Heo, "Application of Artificial Intelligence in Acute Ischemic Stroke: A Scoping Review," Neurointervention 20, no. 1 (2025): https://doi.org/10.5469/neuroint.2025.00052.
- Omar M. Al-Janabi et al., "Current Stroke Solutions Using Artificial Intelligence: A Review of the Literature," Brain Sciences 14, no. 12 (2024): https://doi.org/10.3390/brainsci14121182.
- "Mercy Data Supports FDA Authorization of AI Tool to Detect Life-Threatening Sepsis," Mercy, June 28, 2024, https://www.mercy.net/newsroom/2024-06-28/mercy-data-supports-fda-authorization-of-ai-tool-to-detect-life-/.
- "Predictive Analytics for Patient Readmission Rates," IT Medical, September 15, 2025, https://itmedical.com/ai-predictive-analytics-for-patient-readmission-rates-it-medical/.
- "An AI Tool Detected Structural Heart Disease in Adults Using a Smartwatch," American Heart Association, November 3, 2025, https://newsroom.heart.org/news/an-ai-tool-detected-structural-heart-disease-in-adults-using-a-smartwatch.
- "Provaria: Transforming Care With AI Patient Message Management," Providence, August 28, 2025, https://blog.providence.org/blog/provaria-transforming-care-with-ai-patient-message-management.
- Julie Minda, "Truveta Optimizes Use of Data, Artificial Intelligence and Human Ingenuity to Fuel Medical Research," Catholic Health World, July 18, 2024, https://www.chausa.org/news-and-publications/publications/catholic-health-world/archives/july-2024/truveta-optimizes-use-of-data-artificial-intelligence-and-human-ingenuity-to-fuel-medical-research.