AI in medicine and business needs proof

• In 2026, spending on AI could reach $2.52 trillion, but most companies do not see a clear effect from implementation.
• In medicine, voice systems promise to ease some of the workload, but their use depends on patient trust and performance in real-world conditions.
• Studies do not confirm claims that ChatGPT is superior to doctors; in tests, the system missed some urgent cases.

In two sectors — healthcare and corporate use of AI — the main issue is no longer the technology itself, but the verification of its results. In business, spending on artificial intelligence in 2026 could reach $2.52 trillion, but almost 90% of companies already using AI do not see a measurable impact on productivity or employment. In medicine, voice systems can help with the workload on clinics, but their adoption depends on patient trust and how well they perform in real-world conditions.

In the corporate sector, researchers and consultants point to a gap between implementation and proof of effect. In a PwC report, 56% of executives said their companies had seen neither revenue growth nor cost reduction from AI. In a CloudBees report, it is noted that only 31% of AI spending can be tied to concrete business outcomes, although 51% of leaders are confident in their ability to measure ROI. Gartner also reports that only 28% of AI use cases fully achieved the expected ROI, while 20% ended in failure.

It is also emphasized separately that technical metrics such as accuracy or recall do not always show business value. To reduce risks, companies are advised to define success metrics in advance, conduct independent verification, and tie further funding to confirmed results.

In medicine, the AI voice agent market is projected to grow from about $468 million in 2024 to more than $3 billion by 2030. At the same time, clinics face staff shortages, queues, and a heavy administrative burden. According to the AMA, doctors spend only 27 hours a week in direct contact with patients out of an almost 58-hour workweek. An MGMA survey showed that 59% of practices receive more than 300 incoming calls per day, and more than a third lose 11% or more of calls during peak hours.

Research also points to a trust problem. Patients may be skeptical of voice systems because of spam calls, robots, and weak chatbots. For such services, a short response delay, specialized modules for different tasks, clear disclosure that the system is AI, and personalization based on previous requests are considered important. Separate studies show that people rate advice labeled as AI worse than advice presented as human.

Against this backdrop, Mark Andreessen’s claim that ChatGPT is supposedly better than 99% of doctors also drew attention. However, the studies cited do not confirm this. In a 2024 study, patients considered the chatbot’s answers useful, but specialists pointed to potentially dangerous errors. In a Nature Medicine study in February 2026, ChatGPT Health failed to refer 51.6% of real emergency cases to the emergency department and instead recommended a routine appointment.

Tags: Technology/AI/Well-being/Research