Well, that's kind of how things work now, which is why your physician sometimes has to check with your insurance carrier to see if they'll authorize work. For e

Well, that's kind of how things work now, which is why your physician sometimes has to check with your insurance carrier to see if they'll authorize work. For example, I recently saw a sports medicine doctor for what we feared was a supraspinatus tear. To diagnose that, we needed to order an MRI. My carrier signed off on the imaging, but it turns out that one of the other big carriers will refuse it as a matter of course until an x-ray has been performed. Now, my doctor has forty years of experience, and if she order an MRI it's because she already knows an x-ray won't return any useful information, and will just waste time, money and her patient's radiation exposure limit. But, statistically speaking, the carrier in question has obviously determined that they'll save money if they force providers to route through x-rays regardless. In other words, my provider's judgement was overridden by some Excel spreadsheet somewhere -- not even a Watson or Big Blue. Meanwhile, doctors are being forced to handle more data entry functions, something they understandably hate, and which has probably turned more providers against the Affordable Care Act than anything else. (The provisions are, IMHO, unduly onerous even as I understand the anti-fraud rationale behind their introduction, but that's another issue for another time.) Having said that, given the stringent requirements FDA places on med-tech, I don't anticipate AI providing more than a support role over the next few decades. FDA is always going to want a credentialed human in the critical chain between patient and procedure.