The doctor will see you now, and AI will be assisting. That’s not a distant promise; it’s our present reality. Almost two-thirds of 1,081 clinicians who responded to an American Medical Association (AMA) survey said they recognize the advantages of using AI in their work, and 38 percent said they were already using it at work. The way we experience health care is undergoing a dramatic transformation, leading to faster diagnoses, more personalized treatments, and ultimately better outcomes for patients.
From predictive algorithms that enhance diagnosis accuracy to sophisticated data analytics that speed up drug discovery, AI has the potential to become a pivotal partner to doctors. Today it’s amplifying the capabilities of health-care professionals as well as ushering in a new era of medical breakthroughs and improving patient outcomes. To better understand how doctors are already using AI, we investigated how clinicians and health-care experts are using these systems now, and how it could change their practice today and in the future.
Easing the administrative burden
Health-care worker burnout has become an epidemic. In 2022, U.S. Surgeon General Dr. Vivek Murthy issued an advisory highlighting the urgent need to fix the clinician burnout crisis. He projected a shortage of more than 3 million essential low-wage health workers in the next five years and nearly 140,000 physicians in the next decade. According to a Mayo Clinic study, more than 3 in 5 physicians reported at least one symptom of burnout in 2021. Against the rising tide of care costs and underinsured patients who wait too long to get checked out, there is too little time and too much care needed to make this system work. And caregivers—administrators, nurses, assistants, and clinicians—all put their own health on the line to be there for their patients and communities.
“We recently conducted a study with the Harris Poll that found clinicians spend nearly 28 hours per week on administrative tasks, while medical office staff and claims staff spend 34 and 36 hours, respectively. This includes maintaining detailed patient records, completing insurance forms and referrals, documenting procedures performed, organizing documentation for claims, and inputting claim information into the system,” says Aashima Gupta, global leader of Healthcare Industry Solutions at Google Cloud. Prior to joining Google, Gupta led Digital Health Incubations at Kaiser Permanente, where she brought digital transformation to outdated, paper-based health-care systems. Gupta says health-care systems are already integrating AI into four major areas: transformational care delivery for patients, operational efficiencies, research and development, and a digital “front door” to more effectively triage patients and address access issues.
Clinicians are using AI software today to change how they provide care, creating efficiencies for doctors and quicker answers and care for patients. Instead of clicking hundreds of times to get to an answer, doctors are using natural language processing to ask and answer patient-related questions.
The rise of artificial intelligence in health care presents incredible opportunities to improve patient care. However, the use of AI with sensitive patient data raises crucial questions about privacy, confidentiality, and compliance. Ensuring strong data privacy and security is paramount. This requires a multifaceted approach that includes secure data storage, robust access controls to sensitive patient data, and comprehensive privacy compliance. Google Cloud customers retain control over their data, and in health-care settings, access and use of patient data is protected by Google Cloud’s reliable infrastructure and secure data storage that support HIPAA compliance, along with each customer’s security, privacy controls, and processes.
Beyond searching and asking questions, Gupta says, AI is helping doctors find information within electronic health records (EHRs) to understand the context of a patient’s condition and health status and then perform tasks based on that information. Imagine a doctor asking an AI tool, “Can you generate a timeline of this patient’s care, highlighting key events and progression?” or “Can you transcribe my conversation with the last patient?” Some clinicians use human scribes, who sit in the room with them during clinical visits to transcribe conversations, but not every health system has the budget or ability to hire a clinical scribe. Clinical scribes also raise some concerns for patient confidentiality; some female patients may hesitate to allow a male scribe to sit in on their OB-GYN appointment, for example, whereas an AI tool may not provoke the same discomfort.
The burden of these manual tasks often takes a lot of time, forcing doctors to add notes to EHRs outside their work hours—what Gupta says some doctors refer to as “pajama time.” An August 2024 report by the AMA found that roughly 21 percent of physicians said they spend more than eight hours on the EHR outside of normal work hours on weekdays. AI can help with EHR tasks, leaving doctors with more time to have human engagements with their patients, and their own family and friends, reducing their risk of burnout.
AI is a machine, but ironically, it’s one that can be used to foster more empathy among clinicians. By having an AI agent or software help them with documentation, searching, and summarization, doctors have more space to empathize with their patients—a big reason many came into the medical field in the first place. “We believe AI’s role is to strengthen the relationship between a physician and a patient and to reduce that burnout or pajama time,” Gupta says.
AI can also be used to reduce the amount of manual, repetitive tasks involved with handoffs between nurses from one shift to the next. A Google pilot program at two HCA Healthcare hospitals—TriStar Hendersonville Medical Center outside of Nashville, and UCF Lake Nona Hospital in Orlando—is using Google’s generative AI technology on mobile phones to track nurses’ 12-hour shifts and create summary and task lists. At the end of their shift, a nurse can review the summary, make any necessary edits, and transfer that information to the next nurse on duty. Typically, this process is manual and time intensive and can result in poor patient health outcomes due to information that’s lost in translation or not shared from one shift to another.
Enhancing diagnostic precision
AI is also significantly transforming radiology and imaging, enhancing the work of clinicians. According to Gupta, the majority of all health-care data is images X-rays, CT Scans, MRIs—and AI can be used to maximally process those images and make sense of them.
Recent developments have enhanced AI’s ability to perceive sensory information with more accuracy, which has big implications for imaging and radiology. In the past, tasks such as reading and interpreting the results of imagery from scans and test results could be done only by humans. But AI is now capable of performing these tasks with increasing accuracy. Traditionally, radiologists visually assess X-rays and report their findings on their own or along with the second opinion of another doctor. MRIs and X-rays aren’t black and white—they’re gray and shadowy and can be tricky to interpret. Sometimes the difference between ordering a biopsy and giving the all clear to a patient is in assessing a millimeter-wide edge of fuzzy gray pixels on a screen. AI provides quantitative assessments of images, assisting radiologists in making more precise diagnoses of conditions like broken bones, fractures, and tears.
In its radiotherapy research partnership with Mayo Clinic, Google Health developed an algorithm to help doctors differentiate between cancerous areas and healthy tissue susceptible to radiation damage—a process known as “contouring” that can take a long time when done manually. But when using the algorithm, doctors saw efficiency gains of 30 to 40 percent, Gupta says—meaning faster planning and treatment time for patients, as well as for the doctors helping them access treatment.
Specific fields of medicine like hematology and oncology are data-driven and -intensive. Oncologists and hematologists have a high clinical need for improved, more efficient workflows and better methods for diagnosing and treating patients. Time is of the essence for these doctors; the conditions they treat can spiral out of control quickly if left undiagnosed, and every day counts. However, because society has increased our capabilities to diagnose and treat cancer, many medical institutions have a massively growing trove of data and increasingly complex clinical workflows.
Google’s partnership with New Jersey–based Hackensack Meridian Health offers another glimpse into how AI can be used to aid clinicians in screening and diagnostics. Using AI modeling, doctors at Hackensack are processing large quantities of imaging data and building AI algorithms to predict metastasis in prostate cancer patients. This allows them to provide more accurate treatment plans for these patients.
In 2024, multinational life-science company Bayer and Google Cloud announced they are working together to address radiologist burnout and support more efficient diagnoses with generative AI. Bayer is working to accelerate the development and deployment of AI-powered health-care applications with a focus on radiology, using Google Cloud’s technology—including generative AI tools. Medical imaging makes up about 90 percent of all health-care data. By helping others overcome the challenges of building compliant AI-powered medical imaging software, Bayer and Google can pave the way for lighter, AI-assisted workloads for radiologists.
Humanizing patient care
Personalization in medicine takes many forms, and AI is already facilitating personalized communications with patients and subsequent treatment. Today, health-care systems are using AI to improve population health and close gaps in care by reaching out to at-risk groups with personalized messaging. For example, a health system can identify all female patients ages 45 to 55 and send them reminders to schedule their annual mammogram. But AI allows this outreach to be personalized by connecting it to individual patient data. This approach moves beyond simply scheduling screening tests and instead focuses on the whole patient, recognizing their individual needs and risk factors. For a subset of patients that have a family history of breast cancer, for example, the invitation to schedule a scan could reference that family history to encourage and normalize scans as a regular activity the women in their family do. “You’re not just seeing an X-ray or doing a mammogram—you’re seeing a patient as a whole,” Gupta says.
In India, which has the highest incidence of tuberculosis in the world, India’s Apollo Radiology International is using Google’s AI systems that enhance early detection of TB based on chest X-ray scans to help screen and detect tuberculosis in patients. “These are the areas where AI is going to democratize access,” Gupta says. Over the next decade, Apollo will use Google’s AI-powered screening models to provide 3 million free screenings for tuberculosis, lung cancer, and breast cancer, providing Indians with more timely diagnosis and care.
Gupta says AI can also be used to identify gaps in care, too. “We’ve built EHRs for the past 10 to 15 years, and the U.S. alone has spent $32 billion to create this massive trove of information. Sometimes that information is structured, but often the clinical notes where a patient shares pertinent information that can’t go anywhere else is unstructured,” Gupta says. For example, a patient may share that they are having challenges with housing, which means they may have trouble getting to appointments or having a place to store medicine, which in turn will affect health outcomes. There isn’t an EHR record for those notes, but doctors can use AI to leverage this wealth of information to help meet the unique needs of patients.
AI is also being used to fill the gaps in medical research and treatment. “Traditionally, many medical studies have underrepresented minority populations. Using AI, you can look into datasets and see where the gaps are in the research,” Gupta says. AI models excel at recognizing complex patterns and correlations within data.
At Google, bioethicists, AI researchers, health equity researchers, and clinicians have collaborated to build a framework aimed at reducing biases in AI systems. The framework, known as Health Equity Assessment of Machine Learning performance (HEAL), is designed to prevent AI technologies from exacerbating existing health inequalities, particularly among groups of people who historically experience poorer outcomes. In one application of HEAL, researchers tested an AI dermatology model, and although the AI model performed well in detecting melanoma across most demographic groups, it fell short for noncancerous conditions like eczema, particularly in older adults.
Much like the sophistication of AI technology itself, the way health-care systems are using AI is constantly changing. To make sure physicians don’t feel alienated by these changes, experts insist on transparency with doctors, patients, and new technologies. “Transparency is going to be important,” Gupta says. “Health care moves at the speed of trust, and we want to ensure any transformation within health-care systems is being done with nurses and clinicians and not to them. That’s a humility we need to bring.”
