Run openadapt quickstart --break-it and you’ll watch a fake clinic app with synthetic patients report a save that its server rejected. OpenAdapt reads the record through a separate interface and finds no saved note, so it stops at the Save step and asks a person to check the record. The demo takes a few minutes and keeps its files on your computer. Run it You need Python 3.10, 3.11, or 3.12, which is what openadapt 1.16.0 supports. ...
Say your team enters a referral in the EMR, and then someone posts the charge in a separate billing app. Posting the charge needs one fact from intake, the patient record the referral was saved to. If that fact comes from whatever is on the screen, such as a window title or the last chart someone had open, the charge can land on the wrong patient. OpenAdapt connects the two steps with one named fact. You record intake in the EMR and posting in billing as two separate automations. Posting starts only after intake is done and checked, which means OpenAdapt read the saved record back from the EMR and confirmed it. Posting then starts with the patient ID that record check confirmed. If the ID is missing, the run stops before anything happens in billing. ...
Most automation decides that a save worked by looking for the green “Saved” banner. We tested how far that banner can be trusted. In our fault test, 72 of 90 runs left the record wrong. A check that trusted the banner passed 54 of 72 bad saves (75.0%). A check that read the saved record back by a separate route passed 9 of 72 bad saves (12.5%). A bad save that reports success does more harm than a crash, because nobody gets paged. The error turns up later with whoever owns the record, as a note in the wrong patient’s chart or a payment on the wrong loan. The automation’s check confirmed that something was saved, without asking whose record it was. ...
On the public OpenEMR demo, a compiled replay finished 19 of 20 runs of an 18-step charting task and made no model calls. A computer-use agent finished 10 of 10 and spent about $0.55 a run on model fees. The agent had the higher success rate. On the one compiled miss, the program clicked Save, saw that the screen didn’t change as recorded, and stopped instead of reporting success. Compiled replay was also 1.8x faster at the median. ...
A computer-use agent is an AI model that works from screenshots and chooses each click as it goes. It doesn’t need selectors, an application API, or a recording, so it can take on a task that nobody has automated yet. The cost shows up when the task repeats. The agent calls the model at each step of each run, even when the screens and the goal haven’t changed since yesterday. ...