Health insurer Humana announced on February 3, 2026 that it had rolled out a new AI system called Agent Assist, built with Google Cloud, to support more than 20,000 member advocates who together field roughly 80 million calls a year from Humana’s Medicare Advantage and other members. Rather than replacing the humans who answer the phone, the tool is designed to work in the background during live calls, a distinction Humana has repeatedly emphasized as the health insurance industry faces growing scrutiny over AI’s role in decisions that affect patient care.
What Agent Assist Actually Does
Built on Google Cloud’s Vertex AI platform along with Gemini and Gemini Enterprise for Customer Experience, Agent Assist listens to a call in progress and generates real-time summaries, anticipates what a member is likely to need next, and surfaces relevant account and benefits information for the advocate without the advocate having to search for it manually. The tool also provides compliance guidance during the call and automates post-call documentation, a task that has traditionally eaten into the time advocates have available for the next call. Humana said advocates began using an early version of the tool in October 2025, with a fuller rollout across its member service centers continuing through 2026.
Executives Frame It as Augmentation, Not Replacement
Japan Mehta, Humana’s chief information officer, said the tool puts what the company calls responsible innovation directly into advocates’ hands, allowing them to focus on what matters most during a member interaction rather than on administrative overhead. Chris Sakalosky, a Google Cloud vice president who worked on the deployment, has described the partnership as an example of agentic AI being deployed inside an existing human workflow rather than around it. Humana has stressed that a human remains in the loop throughout every interaction, meaning advocates, not the AI, are accountable for what gets communicated to members and for any decisions that affect their coverage.
Why the Human-in-the-Loop Framing Matters Right Now
The emphasis on human oversight is not incidental. Health insurance AI faced significant new legal scrutiny throughout 2026, as multiple states enacted laws restricting how insurers can use automated systems in coverage decisions. Georgia’s SB 444, signed May 5, 2026, prohibits insurers from basing coverage decisions solely on AI output, requiring a qualified human reviewer in every determination. Illinois passed a related law, the Transparency in Downcoding Act, targeting algorithms that alter submitted medical billing codes without human review. Any AI tool insurers deploy in 2026, even one aimed only at customer service rather than coverage determinations, is being built and marketed with that regulatory backdrop in mind.
The Industry Context: AI Spending Is Rising Fast
Humana’s rollout is part of a broader surge in AI investment across health insurers. UnitedHealth has said it is spending roughly 1.5 billion dollars on AI in 2026, including a digital prior-authorization tool the company says processes claims with a 96% approval rate. Industry-wide, automated systems now handle an estimated 50% to 60% of insurance claims processing, cutting handling costs by an estimated 25% to 40%, according to insurance-technology researchers. That efficiency push has not been universally welcomed: a 2024 American Medical Association survey found 61% of physicians feared payer-side AI was contributing to more prior-authorization denials, a concern that has only intensified as insurers expand AI deployment into 2026.
Two Different Reactions to the Same Technology
Consumer advocates have generally drawn a distinction between AI that helps a human advocate serve a member faster, which is what Humana describes Agent Assist as doing, and AI that makes or effectively drives a coverage decision, which is the practice new state laws are trying to constrain. Some policy analysts remain skeptical that the line between the two stays clean in practice, noting that a tool which surfaces particular information to an advocate during a call can still steer the outcome of that call even without formally making the decision itself. Humana has not disclosed independent, third-party data on call outcomes since the rollout, leaving the practical effects of Agent Assist on member experience largely undocumented outside the company’s own materials so far.
What to Watch Next
As Humana continues expanding Agent Assist across its member service centers through the rest of 2026, the more consequential test may come from how state insurance regulators and physician groups respond once the tool has been in place long enough to generate outcome data. With Colorado, Georgia, and Illinois all now regulating some form of algorithmic decision-making in health insurance, other states are expected to introduce similar bills in 2027 legislative sessions, meaning the ground rules for tools like Agent Assist are still very much being written even as insurers deploy them at scale.
