UnitedHealth Group is committing $3 billion to artificial intelligence — including agents that phone doctors' offices to book appointments and read medical charts aloud to nurses driving to patients' homes — in a turnaround bet that the company most associated with algorithmic claim denials can use the same technology to rehabilitate itself.
The scale is striking even for the largest U.S. health insurer. UnitedHealth plans to spend $1.6 billion on AI this year alone, with the remaining $1.4 billion spread into 2027. More than 80% of its 22,000 engineers are already using AI to write code or build agents, and its generative AI chatbot Avery is live for 6.5 million members with expansion to 20.5 million targeted by year-end, according to Bloomberg.
Key Takeaways
- UnitedHealth Group committed $3 billion to AI agents that call doctors, read charts to nurses, and process claims — a bet that the insurer that drew fury over algorithmic denials can use the same class of technology to restore trust.
- Under new CEO Stephen Hemsley, the company targets a 30% cut in pr.
The company has a trust problem AI won't automatically solve
The investment comes at a moment of acute crisis. CEO Andrew Witty resigned in May 2025 and was replaced by former chief Stephen Hemsley as the company grappled with the murder of insurance division CEO Brian Thompson in December 2024 — a killing that became a flashpoint for public fury at claim-denial practices. A February 2024 investigation by Stat News and ProPublica had already revealed that UnitedHealth used an AI algorithm with an alleged 90% error rate to systematically reject Medicare Advantage claims with little human review.
The political fallout was immediate and lasting. A federal judge allowed a class-action lawsuit over the algorithm to proceed. Connecticut, Texas, Arizona, and Maryland passed or introduced legislation prohibiting insurers from issuing coverage denials based on AI alone. Starting this year, CMS requires payers to provide a specific reason for every AI-assisted denial and publish aggregate data on rejection rates.
Against that backdrop, UnitedHealth's $3 billion announcement is simultaneously a cost story and a reputation story — and the market is watching both.
What the $3B actually buys
The investment splits roughly two-thirds toward enterprise processes and one-third into software products to accelerate Optum Insight's transition to what the company calls an 'AI-first model.' Concrete applications disclosed so far:
Key facts
- AI agents call doctors' offices to schedule appointments in an active trial
- The system listens to millions of customer calls to identify complaint root causes
- AI reads medical chart summaries aloud to home-visit nurses during transit
- 22,000 engineers are building agents for claims processing, fraud flagging, and billing codes
- Avery chatbot handles customer interactions for 6.5 million members today
- AI-driven pharmacy optimization targets a 25% cut in prescription reauthorizations
- UnitedHealth targets 30%-plus reduction in overall prior authorizations by end of 2026
- Morgan Stanley estimates the U.S. healthcare industry spends $80 billion annually on administrative transactions
The company says it is already seeing a 2-to-1 return on AI investment, though it has not disclosed projected net savings or a timeline for the scheduling-agent trial to scale beyond pilot status.
Prior auth: the central battleground
The prior authorization system is where the AI bet is most politically loaded. UnitedHealthcare has pledged to remove prior authorization requirements for 30% of services by late 2026 and to issue decisions within one business day for at least 95% of complete electronic requests. The American Medical Association, which found that 72% of physicians rate UnitedHealth as having 'high' or 'extremely high' prior authorization burden, remains skeptical.
'Using AI-enabled tools to automatically deny more and more needed care is not the reform of prior authorization physicians and patients are calling for,' the AMA said in a statement — a framing that the company's current AI push has not yet resolved.
The industry is moving in the same direction but at smaller scale. Cigna expects to standardize more than 70% of its prior authorization volume for medical services by year-end and has already reduced overall PA volume by around 15%. Humana has deployed Google Cloud's AI Agent Assist for care-support workflows. Neither company has announced a spending commitment close to UnitedHealth's $3 billion.
What to watch
The Q3 earnings call in October is the first moment UnitedHealth will be expected to show measurable results: error rates for the scheduling-agent trial, prior-auth reduction progress toward the 30% target, and whether the Avery chatbot is deflecting calls or generating new complaints. Any state insurance regulator guidance on AI-caller liability — particularly if a scheduling error leads to a missed or delayed treatment — would immediately constrain every insurer's automation roadmap. The CMS aggregate-denial data requirement, due to surface meaningful numbers by early 2027, is the policy signal that will determine whether this $3 billion buy improved care access or just moved the denial process faster.
Source: bloomberg_tech
