AI, Grunt Work, and Training
A soft rebuttal of soft rebuttals to AI’s impact
Scott Rosenberg summarizes a Quartz piece by Sarah Kessler about what Deloitte CEO Cathy Engelbert worries about, but then he goes off the rails, quietly:
When artificial intelligence invades the white-collar ranks, the tasks it will eliminate first are most likely to be grunt-work: the cases that newly minted lawyers must look up, the reports that junior accountants must review. These are unlikely to be tasks that anyone will truly regret handing over to a machine-learning algorithm. But Quartz’s Sarah Kessler raises a valuable question about this kind of shift: If we take these entry-level assignments out of the mix, how will entry-level employees get their training?
It’s a question Deloitte CEO Cathy Engelbert is also asking: Her firm is built on an “apprenticeship model” that involves junior workers building up their sense of judgment on the job by working their way up through a series of increasingly demanding assignments. Deloitte is rolling out AI systems that review contracts way faster than human employees. But AI can’t do everything, and the company still needs middle-level human analysts to deal with the harder issues the AI can’t resolve. How will companies fill those middle ranks if AI is handling everything below them?
Maybe AI will help solve that problem, or train people itself. More likely, middle managers will keep finding new kinds of new grunt-work for newbies to cut their teeth on.
A weak close to a well-constructed start, and I question Rosenberg’s facile utopian attitude.
Where does he get the ‘new kinds of grunt work’ in an article dedicated to saying that AI is already impacting that grunt work? You won’t find that rosy outcome in the close of Kessler’s piece, though. It’s Rosenberg’s unsolicited reaction, not coming from Deloitte’s CEO Engelbert, or Kessler.
Rosenberg’s part of a chorus I hear singing all the time: commenters saying that we’ll find something for the junior accountants and newly-minted lawyers to do, once AI takes over much of what had been used to train them, even while CEOs are wondering, out loud, if that will be true.
Originally posted on stoweboyd.com.