the ConversAItion: Season 1 Episode 5

AI & the Future of Work

In this episode of The ConversAItion, Jim speaks with a renowned expert on digital business, Tom Davenport, about everything from the fears of job displacement to employability in an AI-driven world. A prolific author, senior advisor to Deloitte Analytics and the President's Distinguished Professor of Information Technology and Management at Babson College.
Listen on Apple Podcasts badge
“If you’re open to the idea that your job might be shaped or assisted by AI, you need to learn about how AI works, and be open to learning new skills whenever you can. You need to understand how an AI assistant takes action, and be able to pick up the ball when AI drops it.”
headshot of Tom Davenport

About Tom Davenport

Tom is always thinking about enterprise technology and its impact on the future of work. He regularly writes for publications like the Wall Street Journal, Forbes and Harvard Business Review.  Follow Tom on Twitter @TDav.


Short on time? Here are 4 quick takeaways:

  1. Smart machines are collaborating with humans, not replacing them.

    In 2015, workplace automation was a topic surrounded by fear. Studies reported alarming statistics about large-scale job loss, and instigated a widespread fear of job displacement.

    Over the last five years, the narrative around workplace automation has shifted focus to AI augmentation—the integration of smart machines into the workforce to support and drive successful business results.

  2. As augmentation overtakes the job displacement narrative, AI panic subsides.

    Fear of AI-driven job loss has decreased for a few reasons. For one, businesses today are advocating for AI augmentation, to promote efficient human-machine collaboration. At the same time, the US workforce is not experiencing the magnitude of job loss predicted only five years earlier. And, even for those who do face risk of unemployment, with such a prosperous economy, the prospect of finding a new job isn’t as daunting as it could be.

    But while panic has dwindled, Tom still warns workers against complacency. Based on a global Deloitte survey, businesses are looking to cut costs by automating jobs, but up to 70% said they don’t care whether they re-train their current employees or hire new ones. To encourage employability in the age of AI, workers must proactively decide if they are willing to work alongside AI and evolve along with technology.

  3. A structured production pipeline is key to AI success.

    Analytics are the core of AI, and require an accumulation of data to inform and improve the technology. Companies must implement a well-developed strategy and timeline for AI adoption to augment relevant analytical skills of existing employees and match rapid advances in artificial intelligence.

    To remain competitive, brands must also embed prototypes into business processes, roles and information systems. This will ensure that they pass through an iterative and constructive process of deployment, all while adding value to the business.

  4. Now, AI will create new roles and skill sets. Later, it will revolutionize the workplace.

    In his book The AI Advantage, Tom quotes futurist Roy Amara who said, “new technologies tend to be overestimated in the short-run and underestimated in the long-run.” 

    Right now, humans are experiencing small changes in the workplace. In five years, AI technologies—NLP, deep learning, robotic process automation and more—will be applied in tandem to create tremendous business transformations at scale.

Read the transcript

Jim Freeze Hello! Welcome. I’m Jim Freeze and this is The ConversAItion, a podcast airing viewpoints on the impact of artificial intelligence on business and society. 


The ConversAItion is presented by Interactions, a conversational AI company that builds intelligent virtual assistants capable of human-level communication and understanding.

In this episode, we’ll discuss AI, automation and the future of work. We’re joined by Tom Davenport, a prolific author and well-known business professor focused on analytics and AI innovation. He’s currently the Distinguished Professor of Information Technology and Management at Babson College.

Tom – welcome! And thank you for coming on the show.

Tom Davenport Thanks, Jim! Great to be here.

Jim Freeze There are some folks that work at Interactions who were students of yours and I’m proud to say they are some of our best employees. So shout out to you and Babson.

Tom Davenport Great, we need to get more in there.

Jim Freeze Well we’re hiring, so we’re working on it. [laughing]

Tom Davenport Ok! [laughing]

Jim Freeze To start off, I actually kind of want to dive right into a Harvard Business Review article you write back in 2015. I think it’s one you wrote with co-author, Julia Kirby. In the article, it was titled Beyond Automation, you talked about AI and the fear of job displacement. And that’s obviously–when you’re talking about AI, it’s a very hot topic these days.

At the time—four years ago—you wrote: “suddenly, it seems, people in all walks of life are becoming very concerned about advancing automation.” Could you describe what the atmosphere and anxiety was like around automation in 2015? 

Tom Davenport I think that was just when some people were starting to I think apply excessive amounts of precision to how many jobs might be lost. There was this Oxford study suggesting that 47% of US jobs were automatable and there were a number of other studies related to that. And I don’t think there had been too many pieces advocating what we were advocating. Now, there are more, but this idea of augmentation–that smart humans and smart machines would be more likely to be working together rather than one replacing the other. And so we were advocating that and also saying that it was more likely I think than large scale automation, at least any time soon.

Jim Freeze Yeah and it’s interesting because I want to get to that concept of augmentation. So the framing or the narrative in 2015. How have you seen that change over the course of the past four years? 

Tom Davenport I think that the level of panic might be subsiding a bit, for a variety of reasons. One, I think a number of vendors, one might argue that it could be a little self-interested, but they’re all pushing the augmentation idea. And two, we’re not seeing a lot of job loss. Certainly a lot of the companies that I work with are not laying off people in substantial numbers because of AI. I’m not sure there’s a single one that I know of that falls into that category. Three, the economy’s really great now so it wouldn’t necessarily be tragic if a few people lost their jobs, they could pretty easily find other ones. So I think the level of concern has subsided a bit. 

I just warn people against complacency. I do think we have to think about this issue and we have to think about retraining and re-skilling and what we would do if we did have large-scale job loss and maybe some ways we might prevent it. But, in general, I think people are calming down a bit about it overall.

Jim Freeze Yeah it’s the history of technology for the past five hundred years. There’s technology that comes about and it helps make things more efficient. You know, it can be disruptive and it is disruptive but I don’t know if AI is any different than other significant technologies that have dramatically changed how we operate, how we communicate and how we do our daily work?

Tom Davenport Well, it hasn’t been so far and you know there are lots and lots of AI projects underway in large sophisticated organizations. I think maybe one big difference is that now the potential population affected by this technology, AI, is not the frontline worker, the textile worker or factory worker—it’s the knowledge worker. Those are the people who write books and articles and who appear on podcasts and so on. And so maybe they’re making a bit more fuss about it when it starts to affect professors, and lawyers, and doctors, and journalists and people along those lines. 

Jim Freeze Well I was afraid you were going down the path of me not having a job, so thankfully you didn’t go there.

Tom Davenport Well I did write a piece a couple of years ago about the automation of marketing and as you know, I mean I think there’s still plenty of roles for human marketers. But the same thing that I say about radiology, and journalism, and so on. The only marketers who will lose their jobs are those who refuse to work with AI because that field is being transformed just like every other one, as you well know.

Jim Freeze Absolutely. Let’s come back to this notion of augmentation in that HBR article that I referenced earlier. You suggested that we reframe that threat of automation as an opportunity for augmentation. Could you explain a little bit about the thinking behind that, and what you mean by augmentation?

Tom Davenport Yeah. We really tried to define five different ways that humans could add value to intelligent machines and suggested that most jobs–a lot of jobs would be changed by AI and that the things that humans do, some of them, might be replaced by AI but probably because AI is relatively narrow, entire jobs wouldn’t go away. And we had some recommendations for you know which of those five approaches might make the most sense for different types of workers.

Jim Freeze So do you deal much at all–or provide advice to businesses and industry leaders to promote this notion of augmentation to encourage them to think of it differently, instead of viewing it as a threat, but an opportunity?

Tom Davenport I do consult with some companies and I do educate senior management teams about it but frankly, I don’t see a whole lot of response yet. I don’t see organizations saying augmentation is the approach that we’re going to take, which I think is a good strategy because then your people will be a lot more likely to work with you in adopting AI. And furthermore, I don’t see HR organizations stepping up to say “Okay let’s start getting the workforce ready for these augmentation roles. So I don’t know exactly what they’re waiting for — maybe a little bit more clarity about the exact jobs affected and how they’ll be affected. But at that point it might be a little late to start retraining. And frankly there are some factors that make me wonder how committed they are to augmentation. 

I’m a senior advisor to Deloitte as well and we have done two different surveys of the state of enterprise AI in the US and in a separate global survey, and in the last survey, 63% said they would like to cut costs by automating as many jobs as possible. So that was a bit of a surprise and then a very low percentage – I think only 12% – said they were committed to re-training their current workforce to deal with AI. Much larger number, around 70%, said “Oh you know, we might retrain some. We don’t really care whether we retrain or rehire. And some said they would preferred to rehire. So the level of commitment to the workforce, at least in this sample – you know it’s supposed to be a representative sample of large US company executives – was not terribly high.

Jim Freeze That’s interesting and actually, the point you’re just talking about really ties into the next question I had which is around the advice you provide to, as an example, your MBA students around another notion that you’ve talked about which is employability in the age of automation. What kind of things do you suggest to your MBA students to strengthen their employability in an AI-driven, or soon to be AI-driven, workplace?

Tom Davenport Well I think there’s, you know, one big choice that everybody needs to make: Am I comfortable working with AI, smart machines, whatever you want to call them, as a day-to-day colleague or am I not? And if you are–if you’re open to the idea that your job might be shaped, mediated, assisted by AI, then you need to learn something about how AI works, and you need to be open to learning new skills whenever you can, and need to be trying to understand how an AI assistant makes decisions or takes actions, and being able to intervene, pick up the ball when an AI system drops it. 

Then the other choice is if you say, “No I really don’t want to do that for a living. I don’t want to spend my time at work with a computer by my side doing some of the work.” Then you need to say “Ok, what can I do that machines can’t do?” Or maybe picking something that is so narrow, such a small niche that somebody’s unlikely to be automating it any time soon. And those are hard choices just because you know the technology is changing all the time, getting more capable. We say you’ve got to constantly monitor the state of technology if you’re going to take that strategy of, “I’m going to avoid that stuff.”

Jim Freeze Yeah the latter strategy you just outlined certainly sounds like a risky one.

Tom Davenport I think it is but there are some things that machines aren’t terribly good at today—highly creative jobs, highly empathetic jobs. Or it may be that humans prefer to work with people for certain types of things even if they could be done by machines. You know, maybe we’d rather have a human lawyer or human judge than a machine-based lawyer of judge even if those things are possible. Even if you think about autonomous vehicles, right now people are not very sanguine about having or even driving around in autonomous vehicles. Now I suspect that that will change somewhat over time, but you know it may be that people are willing to pay a little extra for having a human driver – so I don’t know. As you say, that’s a risky assumption but it may end up being true.

Jim Freeze Yeah it’ll be interesting to see. It just seems like a risky proposition long term given how quickly technology moves and how what seems, five or six years ago, not doable is becoming increasingly doable. So interesting for people who go down that path I guess. They’re just going to have to stay on top of or be very narrow in what they choose and hope the need for that exists. Your background is in analytics or you have a background in analytics. Can you explain how the insight that you’ve garnered from data and analytics informs your, your views, your examination of automation and the future of work?

Tom Davenport Well yeah you know one key factor is that analytics and AI are the same in some key respects. I mean, in analytics, we talk about predictive analytics and that turns out to be basically identical to the simpler forms of machine learning, you know which are basically predicting an unknown outcome based on learning from a data set for which the outcome is known. Now machine learning does often have more complex algorithms and I think we’re making a lot of progress at doing things like image recognition and so on, that traditional analytics were not great at, but there are a lot of similarities, so that helps. 

I think a lot of the people who deal with analytics at organizations are fairly well-equipped to learn new skills and start working with AI. So I encourage a lot of companies to you know re-train and augment the skills of their existing analytics people so they can do AI work. Now in the work that your company does, language-oriented applications have historically been kind of a world to themselves and knowing analytics doesn’t necessarily mean that you can do you know an intelligent agent system or even a chatbot really well. There are some statistical versions of them as you know but a lot of them are quite symantec in their orientation and that really comes more from computational linguistics than it does from statistics and analytics historically-speaking. 

Jim Freeze Well it is interesting though that there are capabilities that we have in some of our vertically-focused applications that we could easily characterize as predictive analytics. You know, I’m thinking about, we have a vertical in food services where we’re applying our technology to ordering. And it’s interesting we’ve built some predictive analytics models that suggest what’s the best opportunity for upsell. So there’s certainly some I think applicability as you pointed out and direct applicability of analytics to you know AI. They’re all based on big data, right?

Tom Davenport Yeah and you know I think there are only so many underlying capabilities of AI. There’s statistics, there’s semantics and there’s logic in the form of rules and so on. And basically all AI is some combination of those three things so if you’re aware of those underlying building blocks, then I think you could do a lot of great work in AI.

Jim Freeze I–uh–I was thinking about, I attended a Wall Street Journal event, their first event that they did on AI. They’ve got a focus now on AI and they had an interesting program where they had lots of speakers. The thing that struck me, and it really relates to my next question, is that an awful lot of companies talking about what they’re planning on doing regarding AI, as opposed to what they’re actually doing. And I’m wondering if you have advice for businesses that are thinking about doing AI, know they need to take on some AI initiatives, probably in the context of digital transformation. How do you suggest they move forward with incorporating AI and automation into their organization? Are there best practices? Is there a methodology? What do you suggest?

Tom Davenport Well first I suggest that you shouldn’t wait too long. And I wrote a little piece with a friend and former colleague, Vikram Mahidhar, saying that fast-followers are unlikely to do well with AI because it requires a lot of learning and a lot of data, and accumulating those things tends to take awhile. Some of the other Deloitte surveys I’ve done suggest–I think these must be slightly inflated–but over half have a process for putting prototypes into productions. Almost half have executive champions. 37% have a well-developed strategy for AI. So I wouldn’t wait too long is my first advice. 

And then the other thing that you find you know is companies that are a little bit beyond just planning but they just have a pilot or prototype. They seem unwilling or unable to get that into production and so I’m starting to advise companies to, from the very beginning, and you know it’s okay to start with a pilot, but have kind of a pipeline in mind just as if it were a sales funnel and think about what factors might move it along or what might be preventing it from moving along, have some goals for what percentage of projects move down the funnel successfully and start evaluating your AI people not just on how many machine learning models they produce but how many production of deployments do they achieve and my guess is that would start to change their perspective quite dramatically. 

Unfortunately, both analytics and AI people seem to think that the purpose of their existence in many cases is creating models. And models by themselves don’t add any value to businesses. You have to, you know, embed them into businesses processes and roles and information systems, production systems if they’re going to add any value

Jim Freeze Yeah absolutely, totally agree with you. One final question for you: do you have a kind of audacious prediction for five years from now regarding the state of AI and automation?

Tom Davenport Well I don’t know how audacious it is. I think that a lot of these categories will be combined so it will no longer make any sense to say “Well yeah, that’s a robotic process automation application or a deep learning application or an NLP application.” Because things will all be combined and there will be lots and lots of I don’t know, APIs or whatever all over the place that you can call for whatever purpose you need. 

I think that we’ll start to see—you know I’m a big believer in what I call Amara’s Law. I wrote about that in my recent book on AI called “The AI Advantage.” Futurist Roy Amara said that new technologies tend to be overestimated in the short-run and underestimated in the long-run. And so I don’t know five years is kind of a medium run I think the way many business people think. We’ll start to see I think kind of an impressive collection of things. What you really have to do with AI to really transform an organization is put a number of applications in place and we’ll start to say, “That company really transformed its customer service” or “That company really transformed its product offerings with AI.” Now a lot of it is kind of you know invisible – even at Amazon, Jeff Bezos said that much of what they do with machine learning is “quietly but meaningfully improving core operations.” 

But I think we’ll see some pretty dramatic things start to happen in five years. Some other things like fully autonomous vehicles, I’m still not sure we’ll see those in five years. That may take a little bit longer. But we’ll start to notice the businesses that have really been aggressive about AI.

Jim Freeze Yeah I think you’re right. I think we see that with some of our customers who have really aggressively embraced AI and see the value in it and aren’t doing it as a science experiment but are doing it as you know a major transformation of their business. And they’re garnering real benefits. I mean real benefits from it, whether it be in terms of improved customer experience or whether it be in terms of significant cost being driven out of the business. So Tom, it has been fantastic having you on the show. Really, really appreciate it. Thank you very much.

Tom Davenport Thanks, Jim. I enjoyed it.

Jim Freeze Take care.

Tom Davenport Bye.


This episode of The ConversAItion was recorded at the PRX Podcast Garage in Allston, Massachusetts, and produced by Interactions, a Massachusetts-based conversational AI company. 

That’s a wrap for this episode of The ConversAItion. I’m your host Jim Freeze, signing off, we’ll see you next time.


Check out more episodes of The ConversAItion.