I have seen a few instances of this concept being actually true; but only in (computer assisted) old-and-busted-legacy-education.
I did some contracting on a couple of educational software implementation projects for a school system. One math, one music. In the math case, the software maintained an account for each student and (once told the general level and area to work with) would proceed to pose the student problems, keeping track of accuracy and speed, and adjusting the difficulty of future problems accordingly. It also crunched and pretty-printed the data for the teachers. Fundamentally, nothing that flash cards weren't doing slightly less efficiently since forever, and the software was capable of absolutely nothing useful if a given student was really having difficulty (except making the fact obvious to the teacher so they could do something about it and get a special education person involved if necessary). Within it's scope, though, it was better than flashcards at hitting the 'stuff you need work on; but aren't just beating your head against the wall on' bracket, and it made it easy to ID students with issues, sometimes even the conceptual areas they were particularly weak in, and get them the relevant assistance.
The music one was a bit more sophisticated. It came with a large library of pieces for which it was capable, once given the student's instrument type, of playing any neccessary accompaniment and of recording the student's playing. At it's most basic, this provided an easy mechanism for allocating and collecting 'practice X, Y, and Z for Tuesday' style assignments, since the recordings could be automatically collected, if desired, by the teacher the student's account was associated with. The more sophisticated capability was the ability to analyze the student's play and identify and score the degree of deviation from the correct output across an entire piece. Very neat to watch and also allowed convenient identification of students with weaker or stronger grasp of a piece (Not trivial if you want one music teacher to cover a zillion students. 1-to-1 listening is trivial for a competent music teacher; but finding time to do 90+ sessions of that, at least once a week, while also teaching them something new? Machines have their virtues...) and could show the student (graphically, note by note) their performance on the piece.
In both cases, the software would have been of dubious utility, especially for the hard cases, which it was pretty much only useful for identifying, not remediating; but computers can definitely do good-enough-and-far-more-comprehensive-than-you'd-hire-the-faculty-for high speed analysis of student performance.
I'm unconvinced by their ability to do much short of throwing additional drills at you (barring nontrivial further development of rather hairy problem areas) if you aren't getting it; so both programs would have been a total cock-up without the existing faculty in the loop, except for the strongest students who had the least use for them anyway (vs. almost-as-good flashcards and sitting down for piano practice customs); but very fast feedback was something that they could do, and did do. Probably still do, unless they've let them bitrot...