Healthcare organizations are investing millions of dollars into artificial intelligence to achieve a faster and more accurate care delivery model.

Many systems nationwide are already employing AI for back-office functions such as revenue cycle management, and interest is growing in determining how AI can directly impact patient care and lead to more efficient operations.

Few systems provide details on the scope of their investments. A National Bureau of Economic Research study published in January estimated widespread AI adoption could save the U.S. healthcare industry up to $360 billion a year.

Radiological imaging has become one of the most popular use cases for AI, as the technology is often able to spot abnormalities and detect cancers more quickly than clinicians. However, most providers say they are interested in AI as a tool to augment human intelligence, not replace it.

“There is always some initial skepticism,” said Dr. Howard Chen, chief imaging informatics officer at Ohio’s Cleveland Clinic. “There is a trust that has to be built between the [provider] and the algorithm. … That also helps us to slowly transform our culture from, ‘I don’t know if that AI is actually going to be helpful,’ to, ‘It is a tool, and it is a tool for me to do my job better.’”

Here’s a snapshot of how five healthcare organizations are investing in AI.

Cleveland Clinic

Cleveland Clinic contracts with vendor to incorporate AI into its stroke response. Images from a patient’s CT scan are sent through an AI algorithm that can identify large vessel occlusions, or blockages, in large arteries. If the algorithm detects a blockage, the platform alerts neuroradiologists and connects them with other physicians such as neurologists or neurointerventionalists in a virtual chat room to discuss next steps.

Chen said using AI speeds up the response time, which greatly impacts how badly the brain is damaged. The program focuses on large vessel occlusions because those can be reversed if a patient is treated quickly enough, he said.

The technology can be used to treat brain bleeds, Alzheimer’s and multiple sclerosis, and Cleveland Clinic continues to vet vendors for possible partnerships, Chen said.

Cleveland Clinic also uses iCAD’s ProFound AI program to detect breast cancer in 3D images, a step that has led to more accurate readings, earlier diagnoses and fewer false positives, Chen said. He declined to share specific data on those metrics.

“It also allows our radiologists to read a little bit faster because they now have a tool that actually works,” Chen said. “If it’s in 3D, then your computer algorithm done in the 1980s that ever only saw 2D standard mammography, it has no business performing on a 3D mammography image.”

Chen said Cleveland Clinic has a radiology informatics team of about 60 people who help implement AI across the enterprise. Annual salaries for those jobs nationwide can average anywhere from $60,000 to $100,000 depending on the location, according to online job marketplace ZipRecruiter.

He said Cleveland Clinic evaluates possible AI tools and then assigns a budget to each project if it decides to implement a new capability.

Boston Children’s Hospital

Much of the AI innovation at Boston Children’s Hospital has centered on provider burnout and patient experience, said John Brownstein, chief innovation officer. One example is its use of Nuance DAX, an ambient clinical intelligence solution that records conversations between physicians and patients, freeing up providers to better interact with their patients. Many healthcare organizations use Nuance.

Boston Children’s, which often treats rare and complex diseases, also uses AI and an extensive collection of past radiological images to create an algorithm that can identify abnormalities in new images. Brownstein said the hospital builds many of its AI tools in-house, but it partnered with GE for these radiology use cases.

“We’ve been building an AI roadmap for many years,” he said. “At this moment in time, there’s a lot of excitement, enthusiasm [and] visibility [happening] at a faster pace.”

In intensive care, Boston Children’s uses AI to predict lengths of stay, capacity and whether patients can be extubated, among other thingsBrownstein said.

Funding for AI is part of Boston Children’s larger digital health strategy and has become more of a priority in recent years, Brownstein said. Brownstein noted the challenges of using AI to treat pediatric patients, including the additional consent required and creation of algorithms specifically geared toward children.

Providence Health Services

Renton, Washington-based Providence has a goal to devote at least 5% of its technology and applications budgets to large language models and AI by the end of the year. The system expects to spend roughly $10 million on AI projects this year, said B.J. Moore, chief information officer.

“We didn’t provide any additional funding. I basically said re-prioritize your work. … Health systems are losing money, so I couldn’t find 5% if I wanted to,” Moore said.

Moore said the launch of OpenAI’s ChatGPT several months ago encouraged Providence to make moves in AI. In May, the health system rolled out three AI capabilities built in-house to address administrative burdens: a patient-facing chatbot named Grace, an inbox management program to categorize physicians’ emails based on content, and a platform for physician education and referrals called MedPearl.

Moore said the health system thinks of AI as a co-pilot. It is testing and implementing lower-risk AI tools for administrative purposes and will slowly move into clinical uses—a transition he doesn’t expect to happen this year.

“It’s one more assistant in the room that you don’t have to pay for, and that assistant is super smart, but ultimately the nurse or doctor is going to have to review the recommendation,” Moore said.


This year, Altamonte Springs, Florida-based AdventHealth is piloting AI tools from vendors such as Mednition and Bayesian Health to predict sepsis, or when a patient’s immune system has an overwhelming and potentially dangerous response to an infection. The AI capability, which can assess dozens of variables at a time, tracks patient vitals to detect sepsis risk earlier and allow clinicians to respond more quickly.

“A very small change in temperature or a very small trending in blood pressure can help make that prediction where that’s a distinction that as humans we wouldn’t make that, or it wouldn’t be quite that obvious on our checklist,” said Rob Purinton, vice president of analytics and performance improvement.

Purinton said AdventHealth’s first choice is to adopt AI tools through its Epic software, but it will vet other options, including non-AI tools, if there are better possible solutions for the health system.

AdventHealth’s AI budget is woven into its larger technology strategy, and Purinton estimated the system spends millions of dollars each year on AI, including back-office solutions. He noted AdventHealth uses more than 40 AI-based tools, with another 20 or 30 under evaluation. The system launched an AI advisory board last year, bringing together clinical leaders to educate them on AI tools.

“As a health system, the sort of the challenge we have is, ‘How do we navigate all of these startups, tools and alternatives that are out there when they are becoming so diverse?’” Purinton said. “Our approach is to try to balance forward progress with being prudent about what we implement and making sure that we really kick the tires on any tools before it impacts patient care.”

Banner-University Medical Center Phoenix

Banner Health’s University Medical Center in Phoenix launched a capability in late March with to help detect pulmonary embolisms, or blood clots in the lungs, in CT scans.

The AI tool can assess a patient image within a couple of minutes and if needed, immediately alert the medical center’s pulmonary embolism response team. A multidisciplinary team, which can include emergency physicians, critical care doctors, radiologists and cardiologists, then coordinates a treatment plan on the chat platform—all within a matter of minutes, said Dr. Suresh Uppalapu, a critical care intensivist.

Uppalapu said the medical center chose to work with a vendor because it doesn’t have the capability to build its own tool in-house. He said there are no plans to roll out the pulmonary embolism tool to other Banner locations at this time, as the pilot is in the early stages.

To fund the launch, the medical center landed a $30,000 Highest and Best Use grant award from Banner, which has already worked with on stroke care, Uppalapu said.