UnitedHealth Group told investors on its second-quarter 2026 earnings call in July that authorizations submitted through its new pharmacy platform, PreCheck Prior Authorization, hit a 96 percent first-submission approval rate, with denials tied to missing information down 68 percent and appeals down 88 percent compared to the manual baseline the company ran before. Combined with OptumReal, a separate AI-first claims and coverage platform launched in October 2025 that had already processed roughly 500 million transactions by mid-2026, the numbers describe a company that has quietly rebuilt the machinery insurance runs on, and is now betting its earnings story on it.
The scale of the build-out
UnitedHealth is spending approximately $1.5 billion on AI in 2026, funneled largely through its Optum Insight division under CEO Sandeep Dadlani. PreCheck compressed pharmacy prior-authorization approval times from over eight hours to under 30 seconds in some cases, according to the company, and was deployed in partnership with Cleveland Clinic. A separate tool, Digital Auth Complete, alongside InterQual Auth Accelerator, automates clinical decision support so that, in some cases, providers can receive an authorization decision before a patient even arrives for an appointment. Nearly 95 percent of prior-authorization requests now arrive electronically, about half process in real time, and more than 90 percent clear within a single business day, executives said, crediting AI-driven automation across intake, triage, and adjudication.
How prior authorization got this contentious
Prior authorization, the requirement that a patient’s insurer sign off on a treatment or procedure before it happens, has been the single most reviled friction point in American healthcare for years, blamed by physicians for delayed cancer treatment, abandoned prescriptions, and administrative burnout. UnitedHealthcare has separately announced it will eliminate prior-authorization requirements entirely for roughly 1,700 services, about 30 percent of its volume, by the end of 2026, alongside an expanded ‘gold card’ program that exempts high-performing providers from routine review. AI is being sold as the tool that finally makes the system fast enough that patients stop noticing it, rather than the tool that automates more denials at greater scale.
The physician backlash
That framing is not universal. The American Medical Association has published research concluding that AI is, in practice, leading to more prior-authorization denials rather than fewer, with more than 60 percent of doctors surveyed saying unregulated AI tools systematically deny patients coverage for medically necessary care. Health Affairs has run commentary describing an ‘AI arms race’ in utilization review, where insurers’ claims-denial algorithms and providers’ appeals algorithms escalate against each other with patients caught in the middle, and warning that speed gains for insurers do not automatically translate into better outcomes for patients whose claims get auto-denied and must then fight an appeal.
Regulators are moving faster than Washington usually does
States, not the federal government, have taken the lead on oversight in 2026. Alabama now requires insurers using AI in prior-authorization decisions to base determinations on a beneficiary’s actual medical history and clinical circumstances, and to certify annually that their AI does not rely on generic group datasets that could discriminate against subscriber populations. Indiana’s HB 1271 prohibits insurers from using AI as the sole basis for downcoding a claim without a healthcare professional’s review, and Arizona’s HB 2175 requires independent provider review before an AI-influenced denial can stand. At the federal level, the Centers for Medicare and Medicaid Services this year launched a demonstration project called WISeR, for Wasteful and Inappropriate Service Reduction Model, testing AI-assisted review of Medicare claims, a project that critics argue could formalize the same AI-driven denial patterns state lawmakers are trying to rein in.
What UnitedHealth says the numbers mean
Company executives, including CFO Wayne DeVeydt, have pointed to the approval-rate and turnaround statistics as evidence AI is trimming medical costs and administrative overhead simultaneously, reporting the AI push helped shave roughly 270 basis points off medical costs in comparisons cited around the July earnings call. From the insurer’s perspective, a 96 percent first-pass approval rate is proof the system is approving legitimate care faster, not proof it is rubber-stamping less scrutiny. Whether that approval rate reflects genuinely faster, fairer adjudication or simply a system tuned to approve routine, low-cost requests instantly while still contesting expensive or borderline claims is exactly the question state regulators and the AMA are trying to force insurers to answer with data, not press releases.
What’s next
Expect more states to follow Alabama, Indiana, and Arizona with their own AI-in-claims legislation through the rest of 2026, and expect insurers to keep publishing efficiency statistics as a hedge against that regulatory momentum. The real test will come from CMS’s WISeR demonstration and from independent audits of denial patterns, rather than insurer-reported approval rates, since an insurer’s incentive is to publicize speed and its disincentive is to publicize which categories of claims its AI is quietly getting better at denying.
Photo: Leeloo The First / PEXELS via Pexels
