AI earns patient trust, but hospitals still hesitate

By June 2026, the evidence pointed to a consumer trust shift large enough to support a new care model. HIT Consultant reported that the ZS Impact Institute’s inaugural 2026 Future of Health Report described a structural shift in patient orientation, finding that approximately 90% of individuals using AI and digital interfaces trusted the generated health insights nearly on par with their own physicians, showing AI was becoming a primary source of routine health information rather than a novelty.

By October, that trust was visible in behavior. EMARKETER said consumers were using AI as a research companion and, in some cases, to directly replace doctor advice, including after appointments when people went to the AI tool to interpret what they heard and ask follow-up questions. Ami Bhatt said she started telemedicine for her patients about seven years before COVID, using digital technologies such as a digital stethoscope, handheld ultrasound, and image sharing to teach patients their baselines. She also recalled that in 2014 one patient played the recording of his heart murmur made with one of the first digital stethoscopes. A March 2026 KFFF survey found many users turned to AI because doctor visits were unaffordable or providers were hard to reach.

The economic case is no longer theoretical because the infrastructure of AI-enabled care is already scaling across chronic disease management, diagnostics, and home monitoring. IndexBox says connected devices such as glucose monitors, portable ECGs, and remote patient monitoring tools are now integral to chronic disease management and telehealth, marking a shift from episodic visits to continuous management; just as important, it projects the market index for these tools to rise from 100 in 2025 to 218 by 2035, an 8.1% CAGR that signals a category large enough to change spending patterns.

There are also direct signs that this shift is already showing up in financial and utilization outcomes. UnitedHealth Group reported Q3 2026 net earnings of $4.11 billion, up from $2.35 billion a year earlier, explicitly driven by AI and home health, while a Harvard paper discussed in analysis found US healthcare spending in 2024 came in at 18% of GDP and $5.3 trillion rather than the 21.2% and $6.3 trillion once forecast, a $977 billion gap in which technology accounted for about 14%, site-of-care shifts saved Medicare $94 billion, and adherent GLP-1 users saw total costs fall 5% despite 10% medical inflation.

The skeptic’s case is that the bottleneck is not model creation but institutional permissioning. In the University of Texas at Austin four-app retrospective, early setup was quick — obtaining cloud storage access took one business day, subdomain setup took three, and enterprise AI licenses and Information Security Office review took one to two business days — yet deployment slowed when teams hit enterprise integration and what the authors called “pathway-discovery burden reflected in gaps and uncertainty about the next responsible office,” including roughly two-week delays after AI-license approval on October 17, 2025 just figuring out what came next.

That same control logic explains why health systems will automate internal work before letting AI act on patients. Intermountain Health uses an AI agent to triage IT incident tickets and an AI-enabled HR front door to resolve about 35,000 employee inquiries without opening a case, yet refusals “kept returning to AI that acts on a patient rather than hands its work to a clinician”; Geetha Manjunath put the validation problem bluntly at TechSparks 2026: “The question is not whether you build your model,” while Sivasubramanian said integration into government systems is “a significant challenge,” especially across “around 150,000 primary health centres.”