| A dismal failure rate
U.S. healthcare continues to pour money into health tech, especially anything with "AI-powered" stamped on the label. Yet, a recent State of AI in Business 2025 report from Massachusetts Institute of Technology (MIT) suggests that 95% of these projects, mostly AI pilots, fail to deliver meaningful value. AI pilots perform well in demos but struggle in the complex, risk-aware, and change-heavy reality of U.S. health systems. Roadblocks and resistance
Health tech is often built to impress value-seeking stakeholders and delight end users, while overlooking the actual gatekeepers. As a result, a configuration may succeed in a pilot but run into roadblocks when it comes to governance and security checks for full deployment. Over time, products accumulate sign-off debt, with opaque models, undocumented changes, ad hoc integrations, and weak monitoring. This makes it almost impossible for legal, quality, and risk teams to approve them. By the time a pilot goes live, teams are already pushing against change-resistant staff. A typical health system juggles EHR upgrades, new quality programs, staffing issues, reimbursement changes, and facility projects all at once. Those who try to support a new pilot and test it are already stretched. Tools that add extra clicks, duplicate work, or force context switching won't make the cut. A better approach
Companies can avoid failures by treating future growth as a foundation, not a later stage. Decision makers need to work backwards from scale, building software architecture and commercial architecture together. Pilots should be designed not just to impress internal stakeholders but to meet governance expectations from the start. Pilots gain adoption when they give a nurse back 20 minutes per shift or let a nephrologist see two more complex patients a day. Relationships are key. When healthcare staff view healthtech vendors as change partners, who back them up for anything or everything, they engage better. — By MedCity Influencer Sanket Patel | 
No comments