I should clarify. Lab errors I speak of are not just on the assay level. Context is vital in many situations that are completely outside of your realm of interaction. That said, even accurate results may be dependent on the interpretation of the particular technologist on duty (especially in the realm of microbiology and cell differentials!)
But the errors I speak of go well beyond what the lab reports. For example. A random cortisol drawn on a patient at the wrong time of day can lead to over or under treatment of a life threatening issue, even if the reported value is completely accurate to the lab's best ability. If that value is reported out of context, without detailed understanding of patient demographics (Icu vs ward), nutritional status, medication status, etc, the interpretation of the data is subject to massive error. This is completely invisible to the lab as a whole. I've seen firsthand what happens when completely accurate lab results are misinterpreted because they were drawn at the wrong time or under the wrong circumstances, and presented as a data point within the patient's chart. Things like unnecessary invasive testing, removal of a healthy adrenal gland, needless pain and suffering on the patients part.
The work you do is invaluable, but it is part of a much larger picture that needs careful eyes on by a multidisciplinary team of trained professionals. Taking that decision making out of a physicians hands, at any level, introduced needless risks because that context is not always evident in a patient's chart. I definitely see machine learning as having a key role in the future of our profession, but as an adjunct, NOT a replacement for clinical interpretation.