How AI is changing healthcare IT
Doctors are adopting AI faster than IT can govern it. Here’s how healthcare IT leaders are keeping up.
While we’re still waiting for AI to find cures for Alzheimer’s or new treatments for cancer, the technology is already hard at work across a range of healthcare providers.
According to surveys by the American Medical Association, more than 80% of physicians use AI in their practices. The most common use: ambient scribes that automatically record and transcribe doctor-patient interactions, giving caregivers back the hours they would have spent manually editing electronic health records (EHRs).
More than 70% of hospitals are using the technology to predict patient health trajectories, facilitate scheduling, or automate billing. Roughly half of radiologists rely on AI to help interpret x-rays and other medical images, and about 1 in 6 clinicians use the technology to inform decisions about patient care.
For IT teams, that’s a lot of new technology to manage, all the while AI is transforming ITSM, too.
“In the past, a typical care team would consist of a physician, a nurse, and a therapist,” says Jeffrey Sturman, a managing partner at WittKieffer, a recruiting and advisory firm for healthcare executives. “Today it’s a physician, nurse, therapist, and IT leaders.”
Here’s how AI adoption is changing how healthcare IT teams operate, the technologies they deploy, and the way these systems are governed.
Read also: The stakes of IT are different in healthcare
Tasks: Automation for the people
When a physician is locked out of the hospital network because he’s lost his password, or a new clinician needs to be onboarded before she can start seeing patients, a simple tech issue can delay how quickly patients get treated. But instead of having to wait for a request, physicians and staff are able to keep care on track as healthcare ITSM evolves from a ticket-driven organization into one that combines self-service and automation with AI-assisted human support.
For example, NHS Christie, the U.K.’s largest single-site cancer hospital, uses Freshservice and Freddy AI Copilot to handle support for more than 4,000 employees. Instead of manually triaging emails, Christie’s IT staff receives AI-generated summaries of each support ticket, allowing them to quickly identify and prioritize the most pressing issues. Freshservice also streamlines the hospital’s procurement process, enabling it to bring new specialists on board more quickly and efficiently.
How McLaren, New Balance & more turn service performance into a competitive edge
AI is taking on a lot of formerly manual IT tasks, from password resets and patch management to application testing and security, says Sturman, a former chief digital information officer for Memorial Healthcare.
“Healthcare providers are using AI to automate many of these workflows, in the same way they’ve automated appointment scheduling and prescription refills,” he says. “We’re seeing more and more algorithms and sophisticated models that help relieve the burden on IT staff.”
Technology: Too many models, not enough time
In addition to maintaining traditional IT systems, healthcare teams are being asked to add AI-enabled medical devices to their networks and integrate AI capabilities into their EHRs. As a result, the job of IT has shifted from keeping core systems running to managing complex data pipelines and evaluating AI models, says Varun Gazala, AI innovation engineer for Kno2, a healthcare communications network.
For example, an LLM that excels at speech recognition, essential for ambient scribing, may be less proficient at diagnosing illnesses or reading x-rays. Worse, these models change so frequently that IT teams need to constantly re-evaluate them as benchmarks change, he adds.
“One model cannot do everything,” he says. “IT people need to be continually re-skilling on the latest AI models, so they know which ones are the best for each use case.”
And as the use of autonomous AI agents expands, troubleshooting problems becomes increasingly challenging, Gazala says.
You can no longer just be a tech jockey.
“Traditional IT monitoring is great at telling whether a server is down, but it can’t tell you if an AI is confused,” he adds. “IT has to monitor the entire AI ‘thought process’ in real time—what prompts were used, which databases the agent decided to query, the logic it used to reach a conclusion. Or the team needs to deploy an ‘evaluator’ model that scans the agent’s output and makes sure it didn’t invent any medical codes or violate HIPAA.”
Teams: Tech know-how isn’t enough
Widespread AI adoption is changing how healthcare IT leaders assemble their teams. While knowledge of AI systems is table stakes, healthcare IT leaders are looking for a range of attributes beyond technical expertise, such as curiosity, flexibility, and empathy, says Tim Crawford, founder of AVOA, a research and advisory service for CIOs.
“You can no longer just be a tech jockey,” he adds. “You need to truly understand the nature of the business and how the technology you’re working with fits into that.”
But the biggest changes may be at the top. CIOs and other technology leaders are being asked to not only incorporate AI seamlessly into their business, but also demystify AI to the rest of the leadership team, says Sturman.
“Healthcare IT leaders need to go across the entire organization like never before,” he says. “You have to be an operator, a clinician, a business person, a technologist, and a forward-thinking innovator, and you have to understand how this all gets played out to the patient.”
Governance: It’s a team sport
In an industry as highly regulated as healthcare, AI governance remains a huge challenge for most organizations, made even more complicated by the prevalence of shadow AI.
While 70% of healthcare organizations have established AI governance committees, only 30% maintain an inventory of all AI tools their organizations use, according to surveys by Censinet, makers of healthcare risk management solutions. That echoes surveys by Freshworks, where nearly three-quarters of midsize businesses admit that their employees are using IT tools without direct oversight.
If an AI agent screws up, blaming the bots isn’t going to get you very far, notes Scott Neve, founder of Ops Intel, which provides AI compliance frameworks for small and midsize businesses.
“When an AI agent closes the wrong ticket, resets the wrong access, or tells a customer something false, the business is accountable, exactly as if a member of staff had done it,” Neve says. “The mistake I see is IT owning the tool while nobody owns the decisions it makes.”
Many healthcare providers are still grappling with which AI tools to adopt and how broadly to deploy them, says Sturman.
“Governance is key,” he adds. “You need to put in guardrails and implement AI in an ethical and appropriate manner. But if you want to stay competitive and technically capable, you also need to place some bets on the technology. If you’re a health provider who says ‘we’re not going to do AI,’ you’ll be left behind.”
