You're apparently missing the point that appraisers and AVMs don't drive the market, AI-enabled or not. They just observe/report what the market participants are doing. The consistency of that observe/report is what I'm referring to.
Speaking of consistency, a calculator or a spreadsheet does not "strive for an artificial level of consistency". They just operate consistently. If there's an error it's occurring at the model specification and calibration stages, not at the operation of the machine itself. Same as what happens in appraisals.
Let's say the AI-enabled AVM or appraisal is consistently 5% high or 5% low, those inaccuracies being measurable via a feedback loop. If a lender or other decision knows they're always 5% high or 5% low they can adjust their decision making accordingly. That's the utility of analytical consistency.
-------------
Serious question: Consistent isn't the same thing as accurate or reasonable; it's just consistent.
So which would you expect to return a more consistent result? 5,000 appraisals performed by 1,000 appraisers or that same 5,000 appraisals performed by a single AI-enabled AVM or AI-enabled appraiser or whatnot?