There is plenty of controversy about where Artificial Intelligence (AI) or Machine Learning (ML) actually work well. Lots of organizations are trying to implement some type of AI/ML solution to either save on labor, or find new insights in their data. The success of these projects seems to be somewhat hit and miss. A few hits noted in the media, and quite a few more misses told anecdotally among friends.
There is a place where I think AI/ML is shining, however, and that's in the medical field. I think image recognition is one of those spaces where ML really does a great job, avoiding the fatigue, boredom, and other factors that lead to inconsistency among humans. Radiology has seen strides using ML, first as a verification tool for diagnosis, and later as the primary tool, reducing the load for humans who only look at those cases warranting additional review.
During this Covid-19 pandemic, there is some triage being done in the UK, based on ML reading lung scans. With some initial resistance all over the world, the rising caseload has made AI/ML systems more attractive to all sorts of health professionals that are struggling to get their work done in a timely manner.
There are certainly concerns over rapid adoption, and likely more and more companies will try to sell systems that aren't as well tested. While there is a lot of potential here, we need good vetting of AI/ML techniques, both by humans, through trials over time, and then later to help train new generations of humans. AI/ML can help, but it can't replace humans completely, and we certainly can't use the skills that humans will use to verify the machine's work.