UnitedHealth's $3 billion AI push has bots calling doctors

UnitedHealth Group is investing $3 billion in artificial intelligence across 2026 and 2027, deploying the technology to automate insurance tasks that still rely on phone calls, faxes and paper. The largest U.S health insurer says it is already seeing a 2-to-1 return as AI speeds up care approvals and digs into customer complaints - a strategy executives have placed at the center of a turnaround after profits collapsed last year.

Wall Street expects the push to boost earnings by cutting expenses. Insurers and medical providers together spend $80 billion a year on administrative transactions, Morgan Stanley analysts led by Erin Wright said in a June research note. "The cost savings potential is clear, particularly for manual, data-intensive processes such as prior authorization," they said, adding that UnitedHealth will also profit from selling AI products and services to other healthcare companies.

How UnitedHealth is using AI

The company has more than a thousand AI uses, 20,000 AI engineers, and 117 large language models available for staff to draw on, according to Sandeep Dadlani, chief executive officer of UnitedHealth's Optum Insight data and technology division. One early system, Optum Real, allows medical providers to check in real time whether a service is covered. It has processed about a billion transactions since it started last year.

Other examples include AI that reads aloud summaries of medical charts as nurses drive to patients' homes, and models that listen to millions of customer calls to find the causes of complaints. One trial even has AI agents calling doctors' offices to schedule appointments for patients. "The CEOs lead it," Dadlani said in an interview at the company's headquarters in the Minneapolis suburbs.

Almost 99% of the company's AI applications are administrative rather than clinical, Dadlani said. An internal review board - including medical ethicists, clinicians, technologists, and privacy and legal experts - clears proposals for new uses. "We are not getting into diagnostic AI," he said.

Public trust and legal headwinds

While investors are bullish on AI's potential, UnitedHealth faces a challenge convincing a skeptical public. A Gallup survey last year found 69% of respondents had little or no trust in businesses to use AI responsibly. Most insured Americans consider cost-control tactics such as prior authorization a burden, and about half have encountered barriers to care including delays and denials, according to a poll from health researcher KFF.

UnitedHealth already faces litigation over its use of algorithms. Plaintiffs in class action cases claim the company and others relied on an algorithm to limit care, including admissions to post-acute care following hospital stays. A recent federal inspector general report found companies using the algorithm, from a UnitedHealth subsidiary called naviHealth, had higher rates of denials. Those denials were almost always overturned when patients appealed. The company said the algorithm does not dictate how long patients can stay in care facilities, and that it is used appropriately in line with regulators' guidance.

Public frustration with the U.S. healthcare system fueled a backlash against insurers after the 2024 killing of UnitedHealth insurance chief Brian Thompson. The company and its rivals, under pressure from Washington, have been removing some prior approval requirements. Tim Noel, who succeeded Thompson as head of UnitedHealth's insurance division, said, "You have to gain trust, earn trust through your actions," adding that changes will "take some time for that to actually be felt by people."

Tracking adoption and avoiding drift

UnitedHealth is tracking whether some workers use AI at least once a day as it pushes to embed the technology into operations. Asked whether the technology has caused job cuts, Dadlani indicated it has not. "We don't have any direct correlation to that," he said. The company has also occasionally pulled back an AI model over unintended behavior. "We have all our alerts firing if something's beginning to drift," Dadlani said.

UnitedHealth says it expects to reduce operating costs by almost $1 billion this year, largely driven by AI. The company this week invited reporters to its headquarters outside Minneapolis to show some of its work - an unusual event for a company that historically limited engagement with press. It was a clear effort to change perceptions. "We only approve using AI," Dadlani said. "We never deny using AI."

Why this matters for healthcare, IT, and development professionals

For technical and operational teams inside large healthcare organizations, UnitedHealth's $3 billion commitment signals that AI automation is moving from pilot programs to core infrastructure at scale. The 2-to-1 return figure suggests the company is measuring and validating these deployments against hard financial metrics, not just experimenting. The heavy focus on administrative workflows - prior authorization, claims status checks, appointment scheduling - means the immediate demand is for systems that bridge between legacy phone-and-fax processes and modern real-time APIs. Professionals working on AI for Healthcare Courses or AI for Customer Support Courses will recognize the same pattern: the technical challenge is less about building new models and more about integrating them safely into existing regulatory and operational environments where an algorithm's denial rate can become a legal liability.