When the uncle of Dr. David Rhew’s co-worker visited the emergency department with severe abdominal pain, he received a scan which picked up a stomach aneurysm. However, he died untreated in the emergency room not long after arrival. Why? His scan was one of many in a large pile waiting to be read by the radiologist. This tragedy sparked a question for Rhew: could the outcome have been different if the hospital was using AI?
“If we were to replace the radiologist with AI it likely would have made no difference, but if we were to use AI to read the scan and flag it as urgent, the radiologist could have intervened and perhaps my co-worker’s uncle would still be alive,” explains Rhew, Microsoft's global chief medical officer and vice president of healthcare, health and life sciences.
For Rhew, situations like this illustrate why it is essential for healthcare providers to explore the potential of AI. “Today, we wait for people to feel unwell and visit a healthcare provider, but clinician shortages, limited resources, inefficient processes and various other barriers mean patients often wait a long time to receive adequate care – or perhaps never get seen,” says Rhew. “By using AI, we can evolve how providers operate to significantly improve the quality of care we deliver.”
AI can quickly analyse vast volumes of healthcare data and evaluate current systems to find operational inefficiencies, bottlenecks in workflows and other issues, says Rhew. “This highlights opportunities to optimise resources, cut costs and boost the productivity and effectiveness of medical professionals, while enhancing the patient experience. When combined with other technologies, AI is already allowing us to address some of healthcare’s biggest challenges.”
Today, AI powers virtual assistants that can answer basic questions for patients, aid medical professionals in analysing clinical data or writing reports, automate administrative tasks, and much more. It can also detect ‘hidden’ diseases. Microsoft, for instance, worked with healthcare organisations in the US to apply AI to retinal scans taken during routine eye examinations to identify chronic conditions.
“We found that 20 per cent of the first 50 diabetes patients screened during a trial in Texas had moderate to severe diabetic retinopathy, which will cause blindness if left untreated,” says Rhew. “Meanwhile, during eye scans in California, we tested participants for five conditions – diabetic retinopathy, hypertensive retinopathy, macular degeneration, glaucoma and retinal detachments – and diagnosed them in 67 per cent of individuals. That’s a pretty dramatic finding.”
In both cases, the affected individuals were referred for further treatment. “AI allowed us to intervene immediately to give these individuals the best chance of preventing the worst outcome,” says Rhew. “Similarly, we can apply AI to routine electrocardiograms to pick up issues beyond heart arrhythmias, such as heart failure and valvular disease, or even predict long-term cardiac mortality. Alternatively, we can use AI to analyse voice biometrics during medical consultations to determine if patients are depressed, stressed or anxious and prompt the clinician to investigate further. Proactively screening for illness and intervening early could reduce the total cost of healthcare in the long term.”
While beneficial, using AI to proactively detect disease has exposed a major issue: there aren’t enough clinicians to handle the increased case load.
AI can help reduce workload burden for clinicians by accumulating patient health data and putting together reports
“Sometimes when we use technology to address one pain point, it uncovers another problem downstream,” says Rhew. “However, AI can help resolve the issue it has created by automating time-consuming manual tasks and processes that prevent medical professionals from focusing on higher-value work.”
For instance, AI can vastly improve risk stratification. “Using AI to analyse scans and data allows us to triage patients and quickly direct them to the optimal care pathway,” says Rhew. “Directing people to the right specialist reduces wait times and maximises the chances of the best outcome. It also allows healthcare providers to optimise resources so they can better serve a greater number of patients.”
AI can also help providers practice at the top of their licence and expand their set of responsibilities. For instance, randomised controlled trials have demonstrated that pharmacist-managed hypertension clinics are highly effective in managing blood pressure for hypertensive patients.
This approach of risk stratification, triage and capacity building is particularly beneficial for rural, remote or low-income communities who struggle to access adequate healthcare. “AI is democratising healthcare access in a way that was previously impossible,” says Rhew.
“Also, some individuals do not have transportation means or sufficient money for medications. AI can help identify those who are eligible for ride-share or financial aid programmes. Addressing social determinants of health can have a major impact on one’s health."
Microsoft and partners have introduced a new initiative to expand access to AI in rural hospitals across the US. “We’re assisting rural hospitals, clinics, and health centres to implement AI and other technologies to address their biggest challenges,” says Rhew. “They typically have small budgets and limited resources, so we’re helping them build technology infrastructure, leverage cloud and cyber protection programmes and apply AI-based tools for various use cases. This will support everything from ambient clinical documentation, revenue cycle management and remote patient monitoring to triage and risk stratification and streamlined referral management. These providers will work together as an interoperable network, jointly covering the communities’ needs.”
While many medical professionals recognise the benefits of using AI, others remain reticent to do so.
“Most fears stem from uncertainty around how AI will be used, the impact it will have on the clinician and patient experience, its potential to replace humans and whether it will be applied ethically and securely,” says Rhew. “It’s crucial to address all these issues to alleviate concerns and encourage widespread adoption. We must measure the clinical, financial and/or operational impacts AI applications have on the healthcare delivery process to demonstrate a positive return on investment (ROI). Concurrently, we need to recognise the major impact implementation plays in evaluating ROI. Well-validated AI with poor implementation can lead to negative results. Lastly, we need to train the workforce to use AI effectively, so it can boost efficiency, productivity and job satisfaction. We also need to develop career programmes where AI creates new jobs instead of taking away old ones.”
AI and other technologies are enabling clinicians to provide virtual healthcare to patients who may be able to easily manage their symptoms themselves from home
To alleviate concerns around data privacy and ethics, organisations should invest in strong cybersecurity solutions and develop responsible AI principles to ensure it works as expected. Developing ethical AI principles is the core goal of the Trustworthy & Responsible AI Network (TRAIN).
“TRAIN is a consortium of healthcare organisations in both Europe and the US, which aims to make high-quality, safe and trustworthy AI tools equally accessible to every healthcare organisation,” says Rhew. “Members work with one another and technology companies such as Microsoft to establish best practices for operationalising responsible AI, and share the cost, time and resources involved with testing and monitoring AI models.”
As the technology matures, Rhew anticipates the industry will move from using AI assistants to creating human-led, AI agent-operated teams. Microsoft and its extensive ecosystem of partners will continue to develop the technology and healthcare-specific solutions necessary to make this a reality.
“Microsoft has already embedded AI capabilities across its product portfolio and built assistants to fulfil basic administrative tasks, analyse data and much more,” he says. “We’re making progress with agentic AI workflows that allow humans empowered with agents to collaborate synchronously and asynchronously.
Rhew cites a multidisciplinary tumour board as an ideal example for showcasing the benefits of human-led, AI agent-operated teams.
“Today, a multidisciplinary team of medical experts must collate lots of data, evaluate it to create possible treatment plans, present them to the patient and choose the best route forward together,” says Rhew. “It’s a time-consuming process, but we could cut it significantly by using AI agents to gather and summarise the information. The board can then use its expertise to analyse the findings and make a final decision. Agents could continue working alongside humans throughout the treatment process, taking care of administrative tasks such as booking appointments. Meanwhile, the medical professionals focus on the complex, higher-value tasks that require a nuanced, human approach.”
Demand for human expertise in the healthcare sector will increase in the coming years, predicts Rhew.
“Humans will always be a vital part of any healthcare system,” he says. “Ultimately, we want to develop patient-centric healthcare systems that operate efficiently. AI agents can support medical professionals by tackling time-consuming tasks required to achieve this goal, but AI agents can’t do it alone. We still need humans to make decisions because they have real-world expertise and experience, as well as the ability to think critically, understand nuances in data and deliver services with empathy.”
While Rhew acknowledges that AI is yet to reach full maturity, he argues it has a vital role to play in healthcare, both today and in the future.
“We often set a very high bar when it comes to trusting new technologies – we want to be certain that new technologies work perfectly or near perfectly every time before we implement them,” says Rhew. “However, healthcare is imperfect and unevenly distributed. To understand the impact AI can have on healthcare today, we need to take a closer look at our baseline. How many people do not have access to care or even the medical information required to determine whether to seek help? How many experience unwarranted variations in care? For these individuals, AI can make an immediate and positive difference in their health. If our goal in healthcare is ‘do no harm,’ then we may find that in some cases not using AI may cause more harm than using imperfect AI. And the good news is that AI will only continue to get better.”
Partner perspectives
We asked selected Microsoft partners how their technology is helping healthcare providers adopt AI technology to increase efficiency and consistency of measurable outcomes across the industry.
The Center for Internet Security and Topcon Healthcare explain how they are using Microsoft cloud and AI technology to help healthcare providers improve cybersecurity, data integration and clinical decision-making.
Mishal Makshood, partner alliance manager of Microsoft Azure at Centre for Internet Security (CIS), says: “Healthcare organisations face rising pressure to strengthen their security and accelerate their cloud adoption with limited resources. The Center for Internet Security (CIS) helps healthcare providers use Microsoft Azure to build secure environments from day one with CIS Hardened Images, virtual machine images pre-hardened to the CIS Benchmarks. They enable IT teams to deploy compliant infrastructure in hours instead of weeks while reducing manual hardening, configuration drift and operational risk. One large US healthcare organisation used CIS Hardened Images to standardise security across hundreds of virtual machines supporting clinical and backend systems, resulting in faster deployments, improved consistency and more time to focus on system availability. This helps clinicians deliver reliable, trusted care.”
Ali Tafreshi, CEO and president at Topcon Healthcare, says: “Topcon Healthcare believes the future of healthcare is connected, data-driven and AI-enabled. Built on Microsoft Azure, our platform applies advanced AI machine learning (ML) to convert ocular and associated clinical data into secure, cloud-scale insights that support clinical decision-making. Through Harmony, a digital health platform, healthcare organisations can ingest, govern and unify data, deploy AI or ML models and integrate results directly into existing clinical and electronic health record workflows. In large health-system screening programmes, this end-to-end capability supports earlier detection of conditions such as diabetic retinopathy, streamlines referral and care coordination, expands access and improves operational efficiency and measurable clinical outcomes.”
Discover more insights in the Spring 2026 issue of Technology Record. Don’t miss out –subscribe for free today and get future issues delivered straight to your inbox.