Amid their experiments, pilots, and investments in artificial intelligence, most companies are still in the early stages of figuring out how AI can improve business performance and how it changes jobs.
For a perspective on where organizations should focus, we reached out to Andrew McAfee, co-director of the MIT Initiative on the Digital Economy and co-founder of Workhelix, a startup that assesses companies’ AI opportunities.
Here are excerpts from our conversation with McAfee, who recently published The Geek Way and co-authored The Second Machine Age, edited for space and clarity:
Companies say that they are seeing gains in productivity at a task level and in some specific roles such as coding and customer service, but many are not seeing gains at a corporate level. The nature of the gains is fractional and distributed, and they evaporate. That is giving businesses pause because they feel like they’re making big investments to prioritize AI…
This is the standard process that happens when you introduce a new category of technology. The best analogy is with electrification. What happened at first when factory owners who previously had a big old steam engine in the basement of their factory heard about electrification, the numbers were pretty clear right away that at the task level it was an improvement. So they swapped out their steam engine, they put in an electric engine in the basement, they said, ‘Hey, we’ve electrified our factory, ‘and there were some gains that went along with that.
What happened then, and it took awhile, was that a bunch of wild-eyed radicals showed up and said, ‘You don’t have to stop at one. We can start distributing these things called electric motors all over the factory and somewhere down the road we can imagine a thing that might look like an assembly line or a conveyor belt or an overhead crane. We can really rethink what the factory is.’ I’ve gone back and looked at the journals from that period. That was considered crazy talk, and there was this debate that went back and forth about how far this electrification thing will go. With hindsight, we know the answer to that.
The point I’m trying to make is that the task-level one-for-one replacement is the easy and obvious and probably necessary stuff to do first. The gains keep coming and mount up when you start re-imagining processes, workflows, entire parts of the organization using this general purpose technology. The era of generative AI is a year-and-a-half old now. It’s really early to expect that deeper re-imagining and restructuring to be in full swing yet, but it’s going to happen relatively quickly.
What’s the biggest change to knowledge work that you anticipate over the next two years?
Knowledge workers are going to start very quickly taking it for granted that we have at our disposal—almost no matter what job we’re doing—an army of clerks to help us do the grunt work, the really unpleasant stuff.
There was an article in the Times last year about a physician practice that was a pioneer of using generative AI for the simple work of capturing the interaction, the discussion between the patient and the physician, and it turned that into a physician’s note for the visit. (Let’s assume they dealt with all the HIPAA considerations and all the guardrails.) One of the doctors quoted in the article said, ‘My notes used to take me two hours at the end of the day, and now they take me 15 minutes.’ We’re giving a physician almost two hours with that light a technology lift. We’re giving them back two hours in their day! That’s an example of an army of clerks doing that low-level work that we don’t add any value to.
We’re also going to have a group of colleagues around us all the time. These are AI colleagues, but they can do things like for me, say, ‘Here’s a draft of a chapter. Can you take 15% out of it? It’s just too long.’ Or ‘Where am I being too wordy?’ Technology is already really, really good at that.
The third thing is we’re going to have a bunch of coaches around us all the time. This is closer to what happened in the customer service area that Erik and his colleagues studied. Those folk had an AI coach telling them, ‘Here’s what you can say at this moment with a customer having this problem to lead them to a happy place instead of a bad place.’
My guess is it’s not going to be too much longer before somebody walking into a job and not having that population of AI clerks, colleagues, and coaches is going to be like an accountant walking into someplace in the 1980s and saying, ‘Wait a minute. Where’s my calculator? What do you want me to do with my job?’
How does management change in an AI-enabled environment?
One thing is the clerk work that’s involved with management is going to go down. Aside from that, what good managers do in well-run organizations is primarily keep the social environment healthy, provide a lot of coaching to people, and help the team understand how it fits into the organization and gets aligned with and accomplishes the goals of the organization. That very, very human work of management—not the clerk work, not the database work of management—is not going to go away.
Read a full transcript of our conversation, including discussion of the areas companies should focus on for the lowest-hanging fruit and greatest potential impact from the use of AI, what AI means for the highest-skilled workers, and how McAfee himself is using AI tools.