UnitedHealth's big AI bet

UnitedHealth is not dabbling in artificial intelligence. It is trying to rewire itself around it. The company is reportedly committing about $3 billion to spread AI across the places where money and medicine meet, from claims processing to care management. It already employs roughly 22,000 software engineers, and executives told STAT that more than 80 percent of them now use AI to write code or build software agents. That is not a pilot project. It is an attempt to automate the operating system of one of the biggest companies in American health care. That scale matters because UnitedHealth is not just another insurer. Through UnitedHealthcare and Optum, it touches billing, benefits, pharmacy, data services, physician groups, and the back-office machinery that decides how care gets documented and paid for. STAT’s reporting suggests the company is aiming AI at exactly those choke points: fraud detection, clinical documentation, billing code selection, claims review, and the auditing of payments. Those are the slow, expensive parts of health care. They are also the places where a hidden software choice can change what a patient gets. UnitedHealth has been moving in this direction for a while. In late 2025, executives described hundreds of AI deployments already in use across the parent company. Some are consumer-facing. UnitedHealthcare rolled out tools like Smart Choice to help members search for in-network doctors using preferences such as distance, language, and appointment availability, while also surfacing cost estimates. Other uses sit deeper in the plumbing, inside call centers and revenue-cycle systems, where the goal is to make workers faster and to standardize decisions that used to be handled one case at a time. That is why this new spending push lands differently. It comes after years of selling health care as a data problem that can be optimized, and after public fights over what happens when optimization reaches clinical decisions. UnitedHealth has already faced a class-action lawsuit over allegations that an algorithm was used to help deny post-acute care for Medicare Advantage patients. A federal judge in Minnesota allowed part of that case to proceed in 2025. The company disputes the claims, but the larger point is simpler than the lawsuit: once AI moves from chatbots to coverage decisions, the stakes stop being abstract. The market, for now, is focused on a different question. Can all this automation actually improve results fast enough to show up in earnings. UnitedHealth is scheduled to report first-quarter 2026 results on April 21. Analysts are watching medical costs, Medicare Advantage margins, and the performance of Optum Health. That makes the AI push more than a technology story. It is also a margin story, because every promise attached to AI inside an insurer eventually cashes out as lower administrative costs, tighter coding, faster claims handling, or fewer human touches. That financial backdrop shifted again this week. On April 6, shares of UnitedHealth jumped in extended trading after the government finalized a 2.48 percent average increase in Medicare Advantage payments for 2027, a much better outcome than the near-flat proposal released in January. The move offered a burst of relief to insurers that had been squeezed by rising care costs. It also bought UnitedHealth a little more room to make its giant software gamble look like discipline instead of desperation. Now the interesting part is no longer whether UnitedHealth believes in AI. It plainly does. The interesting part is where the company is putting it: inside the routines that assign codes, route claims, flag fraud, guide call-center workers, and shape the paperwork that decides who pays. That is where health care becomes real for patients, usually one opaque transaction at a time.