Clalit Health Services has rolled out an artificial intelligence system called AI–PRO, built on the C–Pi platform, to help family doctors practice what it calls proactive and personalized medicine. The machine scans computerized medical records every night, cross-references them with clinical guidelines and knowledge bases, and then surfaces recommendations and patients at risk. The doctor still holds the final switch: the system raises a flag, and the doctor decides whether to contact the patient, change treatment, send for a test, or ignore the recommendation.
The Algorithm Watches, the Doctor Decides
The setup is simple enough to sound harmless and bureaucratic, which is usually how these things enter the room. AI–PRO goes through medical information stored in computerized records overnight, then compares it with clinical guidelines and knowledge bases. Its job is not to replace the doctor, at least not yet, but to sort the population into alerts, risks, and follow-up tasks.
Examples cited by Clalit include a patient with diabetes who has not performed tests, a female patient with unbalanced hypertension, patients at risk for osteoporosis, and medication combinations that require attention. In each case, the system identifies a possible problem and pushes it up to the doctor. The human clinician then decides what happens next.
That division of labor matters. The software does the surveillance; the doctor does the governing. The patient is not asked to participate in the nightly scan of their own record. They are processed, flagged, and then managed through a clinical chain of command that remains intact, just with better software.
Proactive Medicine, Centralized Control
Clalit describes the system as a tool for proactive and personalized medicine. In practice, the article says, it is designed to surface patients at risk and recommendations for doctors to consider. The examples given are routine but revealing: missed diabetes tests, unbalanced hypertension, osteoporosis risk, and medication combinations that need attention.
The language of personalization is doing a lot of work here. What is actually described is a centralized sorting mechanism that combs through records every night and presents a list of people who may need intervention. The doctor remains the gatekeeper, deciding whether to act on the recommendation or leave it alone. The patient remains the object of the process, not its author.
There is no suggestion in the article that the system changes the basic structure of care. It does not hand control to patients, and it does not flatten the hierarchy of the clinic. It adds another layer of institutional oversight, one that can see more, sort faster, and keep the workflow moving.
Who Holds the Levers
The article makes clear that AI–PRO is advisory, not autonomous. That detail is the whole story. The system can flag a patient with diabetes who has not done tests, or a woman with unbalanced hypertension, or someone at risk for osteoporosis, or a medication combination that requires attention. But the decision to contact the patient, alter treatment, order a test, or ignore the alert stays with the doctor.
So the power structure remains familiar: data collection at scale, algorithmic triage, and professional discretion at the top. The machine does not abolish hierarchy; it helps administer it more efficiently. In the clinic, as in every other institution that loves the word innovation, the promise is better service, while the actual arrangement is still top-down management of people through records, flags, and recommendations.
The article offers no grassroots counterpoint, no patient-led model, no horizontal alternative. Just the familiar choreography of institutional medicine, now with an artificial intelligence system watching the charts at night and handing the morning queue to the doctor.