In an opinion article published Wednesday in The Guardian, Gleb Tsipursky, CEO of the future-of-work consultancy Disaster Avoidance Experts and author of “The Psychology of Generative AI Adoption at Work,” wrote that a personal AI tool he built flagged deep vein thrombosis, or DVT, as a possible cause of his persistent calf pain. He said the tool, which uses his medical records, medications, lab work, and visit notes, pointed him to seek an ultrasound scan.
Tsipursky described having what felt like a muscle spasm in his left calf for five days, with tenderness, swelling, and worsening pain. After his chiropractor treated it as a muscle issue and the pain persisted, he consulted his AI tool. It suggested DVT and recommended urgent assessment and ultrasound.
He called his primary care office, which advised him to schedule an appointment or go to urgent care. However, neither his doctor’s office nor urgent care could provide the ultrasound scan. Concerned about the possibility of a pulmonary embolism — a potentially fatal complication when a clot travels to the lungs — Tsipursky said he went to the emergency room despite anticipating a long wait. An ultrasound found four clots in his left leg.
Tsipursky said he later learned that his wife’s grandfather and the mother of one of her close friends had died from pulmonary embolisms.
He wrote that his account “is not an argument for replacing doctors with machines.” The emergency room doctors ordered imaging, interpreted results, decided on treatment, and sent him home on blood thinners. “The AI did not cure me,” he wrote. “It just helped me ask the right question in time.”
He pointed to a study published in Science by researchers affiliated with Harvard Medical School and Beth Israel Deaconess Medical Center that tested a large language model on clinical reasoning tasks, including real emergency department cases. According to Science News, the model was more likely than physicians to include the correct diagnosis among possible answers.
Tsipursky cautioned against patients using unregulated chatbots for medical guidance. A Guardian report cited in the article found that one in seven people in the UK are using AI chatbots for medical advice instead of seeing a general practitioner. He said, “A chatbot cannot examine a leg, hear breathlessness, notice distress or take responsibility for a patient.”
He called for regulation, testing, transparency, and clinical supervision of AI in healthcare, as well as “humility from institutions that too often expect patients to navigate fragmented systems alone.” He argued that a safe AI assistant trained on a patient’s medical data can help patients gather records, assess whether urgent symptoms are being missed, and push for appropriate diagnostic steps.