Round $5 trillion is spent on U.S. well being care every year, a system that many Individuals fee poorly, and one which is able to confront even higher challenges within the years forward as annual bills proceed to rise, the inhabitants ages, and medical workforce shortages persist.
Pharmaceutical firms, hospital programs, insurers, and pharmacy advantages managers function in a posh business that’s earned a status for siloing information and badly lagging different sectors like banking and telecom in terms of deploying new technological developments. Larger investments in synthetic intelligence are lauded as a key therapy for the well being care system’s most critical issues, although most within the business report they nonetheless lack adequate sources and planning in AI and different digital investments.
Sandeep Dadlani, the chief digital and know-how officer at insurance coverage large UnitedHealth Group, says he understands that the system is damaged. “I empathize with all people who has a grievance in the direction of well being care,” says Dadlani. “The system must be fastened by many gamers, not simply by us.”
UnitedHealth, ranked fourth on the Fortune 500, says it’s seeing some large productiveness features and adoption from the corporate’s 1,000 AI use circumstances already in manufacturing. These functions, a mixture of generative and conventional types of AI, embody greater than 60 million traces of AI-written software program code that has already been accepted by round 20,000 engineers, 65 million buyer calls that had been initially answered by an AI chatbot in 2024, and 18 million AI-enabled searches to assist prospects discover a physician that had been carried out within the first quarter of 2025 alone.
One other generative AI software anticipated to launch later this 12 months within the UnitedHealthcare and Optum subsidiary apps is a consumer-facing conversational bot that can assist prospects discover a physician, schedule appointments, or evaluate their lab outcomes. That software was initially examined with UnitedHealth’s workers earlier than it would debut externally.
“It’s a maze,” says Dadlani, of the labyrinth that prospects face when coping with well being care suppliers, insurers, and laboratory companies. “So one thing that helps navigate via that and supplies nice end-to-end entry, that’s the dream.”
Dadlani spoke to Fortune in early Could, earlier than the shock announcement that CEO Andrew Witty had resigned. Dadlani’s efforts overseeing the corporate’s AI evolution come as UnitedHealth, which serves greater than 52 million customers, undergoes a tough interval. In December, UnitedHealthcare CEO Brian Thompson was murdered in Manhattan and in April, the corporate’s first monetary outcomes for the reason that killing, UnitedHealth reported earnings that missed Wall Avenue expectations and reduce its bottom-line goal for the complete 12 months. The corporate can also be reportedly going through a Justice Division probe for attainable prison Medicare fraud.
There’s additionally a class-action lawsuit within the courts that alleges UnitedHealth has used AI to illegally deny claims. Dadlani refutes this. “AI isn’t used to disclaim a declare,” he says. “If a declare isn’t eligible to be authorized, it goes as much as a human agent,” who, he says, then makes the ultimate dedication.
Dadlani says round 90% of claims are auto adjudicated, the method wherein software program is ready to manually evaluate a declare. Of the ten% of claims that undergo an additional step for evaluate, many of the points are clerical, which means there are lacking particulars or the data wasn’t enter into the system correctly. After that guide evaluate, Dadlani says 98% of claims are authorized and for the remaining 2%, denials are usually attributed to both ineligible advantages or due to medical or medical security determinations.
Later this 12 months, Dadlani says he intends to launch new AI merchandise that can make auto approval for claims even larger. “We’re already seeing in our early experimentation that AI may also help fill among the lacking info for these claims,” he says.
UnitedHealth’s huge AI portfolio is reviewed on a month-to-month foundation by each enterprise unit. On a quarterly foundation, chief info officers come collectively to watch how AI is getting used throughout the enterprise. The corporate has additionally established a accountable AI board that could be a mixture of inner and exterior technologists, clinicians, authorized consultants, and others who evaluate lots of of AI use circumstances every month earlier than authorizing what can go into manufacturing.
Dadlani says UnitedHealth displays AI use circumstances for security, bias, equity, and authorized compliance. He stresses that AI isn’t getting used to carry out a medical prognosis. “We don’t see AI changing medical doctors or clinicians,” he says. “We wish AI to be a software.”
To convey the workforce alongside, UnitedHealth launched a complicated AI studying course in early March that noticed greater than 10,000 enrollees. The corporate has additionally arrange a devoted platform, referred to as United AI Studio, that permits workers to securely entry the massive language fashions supplied by massive AI hyperscalers, in addition to the small language fashions that UnitedHealth developed by itself primarily based on proprietary datasets.
The corporate additionally had about half-a-dozen use circumstances in manufacturing primarily based on agentic AI, which is designed to extra autonomously full complicated duties—ideally with restricted or no human supervision. Many of the agentic AI use circumstances have been for repetitive, administrative duties. The know-how might want to mature extra earlier than it expands to different elements of the enterprise.
“We will probably be very cautious after we get into extra medical use circumstances and the usage of brokers,” says Dadlani.
John Kell
Ship ideas or recommendations to CIO Intelligence right here.
This story was initially featured on Fortune.com
