AI is being deployed in narrow clinical roles

A colonoscopy already puts a camera inside the colon. The new twist is software that watches the same video feed in real time and flashes a box when it thinks it sees a polyp, which is a small growth that can turn cancerous later. (fda.gov) That is the job GI Genius was cleared for in the United States in 2021: not diagnosing every disease, and not replacing the doctor, but helping spot possible lesions during the procedure itself. The Food and Drug Administration described it as the first artificial intelligence device authorized to detect colon lesions in real time during colonoscopy. (fda.gov) Medtronic says GI Genius can raise adenoma detection rate by up to 14.4%. An adenoma is a precancerous polyp, so this tool is aimed at one narrow moment where a miss can matter years later. (medtronic.com) Mercy Southeast said on April 8, 2026 that it is now offering AI-assisted colonoscopies, and gastroenterologist Dr. Brent Keller described the system as “another set of eyes” during colorectal cancer screening. The hospital’s announcement focused on small polyps that can be hard to see on a moving video feed. (mercy.net) Radiology is a different bottleneck. A scan can be taken in minutes, but a report still has to be read, typed, checked, and sent back, which is why turnaround time becomes a problem when one specialist covers many hospitals. (damoh.nic.in) That is where the India story fits. Bengaluru-based mlHealth360 said on April 9, 2026 that it signed a memorandum of understanding with the Government of Madhya Pradesh to pilot artificial intelligence-enabled radiology in 10 district hospitals. (biospectrumindia.com) The company and local trade coverage both described the pilot in operational terms: improve diagnostic accuracy, reduce reporting delays, and support faster decisions in public hospitals. The plan is a pilot first, with wider rollout only if the 10-hospital test works. (biospectrumindia.com) (expresshealthcare.in) mlHealth360 describes its products as imaging workflow tools that plug into existing hospital systems, which tells you what kind of artificial intelligence buyers are actually signing for right now. They are buying software for one department, one queue, and one measurable delay. (mlhealth360.com) Put the two cases side by side and the pattern is plain. One tool watches colonoscopy video for polyps, and one tool helps district hospitals move radiology scans through the reporting pipeline; neither one looks like the all-purpose “doctor bot” promised in splashy demos. (mercy.net) (expresshealthcare.in) That is how clinical artificial intelligence is arriving in 2026: less like a robot replacing a hospital floor, and more like a spell-checker inserted into one step of one workflow where missed findings or delayed reports already have a cost. (fda.gov) (mlhealth360.com)