Skills for a Workforce of Humans, Agents, and Robots

0:00This message comes from NPR sponsor Paramount Plus and the new original series The Madison. Taylor Sheridan's most intimate story yet. The Madison follows a family raised in a world of digital distraction, forced by tragedy to truly see one another and come together. Starring Michelle Pfeiffer and Kurt Russell, The Madison, new series streaming now, only on Paramount Plus. Thank you for joining us for another podcast from the Commonwealth Club of California. Well, hello everyone and welcome to this evening's program and a big thank you to the Commonwealth Club for bringing us together on this important topic. My name is Anu Madgavkar. I'm a partner at the McKinsey Global Institute, which is the economics and business research arm of McKinsey and company. And I'm also the co-author of our new report, agents robots and us

1:04skill partnerships in the age of AI, which we'd love to unpack and explore more about this evening. And you can download the report for free using the QR code that you see on the screen. Now tonight, I'll be talking with my McKinsey colleague Alexis Krivkovich, who's a co-author of the report, as well as Katie George of Microsoft, and Kevin Delaney of the San Francisco Standard and Charter about some of the major themes of the report. But first, sharing a sort of brief overview of what the research found and what we put forward. So what exactly is the research about? AI is a topic that sparks reactions ranging from extreme excitement to extreme anxiety in most audiences, excitement for the possibilities, anxiety about the impact on human jobs and work. And both those perspectives are valid. But in the research, we really set out to go beyond the conventional headlines around AI's coming to take our jobs.

2:08And really explore a more perhaps interesting and definitely useful and tractable question, which is how does work really change in the context of AI? And what does that mean for human skills that organizations need and are seeking to build? So rather than putting humans against machines, the main finding and message of the report is that this next frontier of productivity, that AI promises, will not involve replacement alone, but will involve partnership. So work of the future will be increasingly a series of collaborations between human workers, AI agents and AI-powered robots, each bringing to the table the skills and the strengths that are unique and that they do best, but interacting together to deliver and create more. Now make no mistake, the impact of all of this is huge and sort of profound and the transitions

3:15and the extent of change will be painful. And that's really because the scope is vast. So even based on today's proven technologies, we don't need even further technology breakthroughs although they keep happening, but even based on what we know today, we can say that more than half of all work hours that are performed in the US economy today could theoretically be automated. So that's 57%, it split roughly into about 44% that could be automated through AI-powered agents and another 13% or so where robots could actually automate that work. But it's a large amount of potential automation and it's widespread, it exists across industries and sectors, there's no occupation in the economy or no role that will be unaffected by AI, but equally there'll be very few, if at all, any roles that will be completely eliminated by AI. So the big messages that roles and work will change and that change could be

4:19ordered of magnitude anything from 30 to 50% of what we do currently. Now applying the prism of skills to this question of changing work, when we mapped all of the skills that employers currently demand in the workplace to the activities that could potentially get automated or not, we found that the vast majority of skills, some 70% of all skills fall in both buckets, things that we use in work that we think only humans will do, but also we apply the same skills in activities that we believe AI could do. And so this notion of shared skills is actually quite important, it means that human workers will need to upgrade almost super skill their expression of things like problem solving, communication, creativity, empathy, judgment. All of these things, many of these things will be offered by AI and human workers will need to work with AI to kind of upgrade their expression of what that skill

5:26means and then collectively create something that goes beyond what AI can do. And the other implication of skills is that with the notion of all this change in the way we work, a lot of careers will be very fluid, career pathways will be unknown, people will have to keep changing what they do and so the notion of transferable skills, things that we can use in different contexts will become more and more important. But already we see that in the labor market AI fluency is rising to the top of the list as the fastest growing family of skills, we looked at this over a two-year period, demand for AI fluency has risen sevenfold and it's getting wider and deeper in the labor market. So well beyond just technical roles but into business, finance, management, sales, operations, all kinds of other roles, now see job postings where AI fluency is a required skill.

6:27And as this pans out it's going to be worth a lot, there is a lot of economic value at stake. We think about this in terms of the value of human capacity that gets released or freed up or available for us to reuse and repurpose in doing things that are more value added and the value of all of that is almost $3 trillion and it could be realized in the space of five years based on adoption rates that we might expect but it will require a lot of redesigning of workflows. This is not just a surgical removal of tasks through the automation process but reimagining and reinventing how work gets done with AI embedded into those workflows. The final thing I will say is that the productivity potential is high but there are a set of real leadership choices that will dictate or determine how much of that productivity we achieve

7:29and what the business and human outcomes really will be. A range of choices that would go from things like perhaps do I treat this as a wholesale organizational transformation using AI versus thinking about it as a much more contained and narrow and specific IT upgradation like that's one set of choices. There are choices around the pace of how fast you move as an organization. What are those things where we should move fast but what are those things that might benefit from needing to go a little more slowly because you need to build trust or ensure safety for example and crucially what are indeed the kinds of work where humans will remain essential. What are the things that we think machines can do more or less autonomously and what are the things that will benefit from partnership and collaboration between the two. So to unpack more of these questions and many more questions I'm sure I'd like to welcome on to stage our moderator Kevin Delaney along with my fellow panelists Alexis and Katie.

8:39Thank you. As a new mention I'm Kevin Delaney on the editor in Chief of the San Francisco Standard and Charter of Future Work Media and Research Company and as she mentioned also please join me in welcoming Alexis Cripp Covitch senior partner at McKinsey and global co-head in North American leader of McKinsey's people and organizational performance practice. She's also co-author of this report Alexis. Thank you. And we are also joined by Katie George who's a corporate vice president of workforce transformation at Microsoft and of course you've already met a new Mad Gavkar partner at the McKinsey Global Institute. So before we dive into the discussion if you have any questions for our speakers there are cards that are on your seats and if you are joining us online you can fill out questions through the YouTube chat and then you pass the cards over and then we'll leave time at the end for any questions that you might have. So we're going to dive in a new one. I'm going to start with you there are a number of

9:40things that you said that I think really we could unpack a little bit and the first one is you're talking about how 57% I think of work hours in the US today could be automated and what you said is that the way that this could play out is that roles or work could change. Can you give an example of how a role or a job might change as a result of the what you're forecasting with AI and agents and robots? Sure we're actually seeing a lot of real-life examples right and we looked at some of that in the context of the research but take for instance a very standard setup of customer service we have customer service agents who typically you know take in calls and service questions or you know do things that customers need and the big issue with a lot of these outfits is that you you don't have very good customer satisfaction and you also have you know waiting time it's not so efficient and so on so the way such a process is being reimagined and

10:46redefined using AI is to really break it up into things that agents that are designed to do specific parts of of this task can do pretty much autonomously and that includes things like figuring out on an incoming basis what the intent of the call is deciding whether it's something that the agent can handle pretty much autonomously there are many sorts of calls which are relatively easy to do if it's more like a self-service type of call then the agent decides that and then routes it to the back end again in a seamless way and if it's more complicated or might require human interface the agent not only figures that out but also then prepares the inputs or the content or the context will prepare the whole case and then pass it on to a human agent who will then have a much better conversation fewer of those conversations maybe you know talk to only 20% of the cases because that those the cases that need need actual talking but do so on a more

11:48informed basis and get much better customer satisfaction as well as a more efficient streamlined process so this is the notion of both reimagining the end-to-end piece with very autonomous agents working pretty much on their own but also the interface the interaction the collaboration that says he has an agent who's made more effective because he has a human who's made more effective because the agent is around to do so many parts of that and and you're imagining that the human is still there and it's not there's not just like late offer or their job is limited but rather they just do different kinds of work is that there might be different numbers of humans and certainly doing different kinds of work so I think both the number of people as well as what they do okay so there could be fewer people and and serving a sort of higher touch customer service can you mean we've come to you I know at Microsoft year so you're you have this job the super interesting job where you're looking at how the workforce is changing how work is changing both in Microsoft internally and at

12:53some of the frontier companies that you work with and I think you've just done 75 case studies within Microsoft of how AI is playing out can you talk about how what you're seeing inside your own company yeah so as you said my team is learning across all of Microsoft and looking at what are we actually doing to change our own work processes and what I found really interesting is it is true that AI can automate a lot of tasks that humans used to do and you can see that that is a large portion of some some jobs and smaller portion of other jobs but what's far more interesting and honestly what I didn't expect is that the greatest value that AI is creating is actually adding new capability into a work process and so that's not about eliminating people's tasks and creating capacity or just dropping you know labor cost savings to the bottom line it's adding new capability or new quality in so for example our audit function has really leaned forward into AI in a really exciting way and while they would say that they can do an audit now with you know in more

13:57productively what's far more valuable is that AI allows them to do proactive risk identification as part of the audit so that wasn't actually defined as the task you know that you would have been analyzing before it's a whole new set of things that AI is bringing in that is creating far more value than whatever the cost savings are and it is very much a partnership between humans who really understand what value is and how to direct AI and how to quality control it and the AI agents and and other tools so think about the Salesforce preparing for a sales call it's not just that AI is making the sales rep more efficient by doing some of the busy work that they otherwise would have to do is actually that AI is bringing all of this new analysis and perspective and recommendations and practice sessions in so that the quality of the sales call preparation is at a totally different level and so there are tons of those examples far more examples inside Microsoft where AI is augmenting the humans than where AI is primarily just automating things away.

15:04One of the things we know is that a lot of companies are investing in AI now maybe not as much as Microsoft but a number of companies are actually feeling like it's hard to see the value from that and I'm interested you just mentioned the sales example that might be a good example of how you how you actually go from your investment to seeing a new and business benefit or customer benefit or that company. Yeah I think there are two reasons why some companies are struggling with you know really seeing the ROI one is that the ROI we all thought was going to show up as labor cost savings in this very straightforward way and it's not showing up that way it's showing up as quality and speed and better customer experience better employee experience and that it's actually far more strategic value for most companies it certainly is for us but it's actually not something that's easy to measure and connect and so one of the things we're doing is trying to standardize the way we look at AI impact across all of those different metrics because often those metrics

16:08are more important than others one good example of this in our sales force is that we look at revenue per sales rep and we know that the reps with the highest use of the AI tools have the highest revenue per rep so we're seeing that connection between growth and AI usage and so what we're asking is for every one of our leaders to be really clear about what they're trying to achieve what's important to their business what's important to their part of the business and therefore what are the metrics that they're going to look at and how do they gear their AI approaches to achieve and that's actually the second thing that I think we're finding is part of the answer to more consistently create clear business impact from AI we've actually developed a playbook because you know with three different recipes for how do you connect AI solutions into workflows in a way that drives business performance because if you just ask everybody to adopt copilot and you know do their own prompt engineering or design their own agents with copilot studio you don't

17:09necessarily get a lot of things that accrue to clear business performance improvement so there's a real need for leaders to be clear about what they're trying to drive and then to really engage their whole team in redesigning workflows end-to-end in a way that delivers that performance. What you're saying about AI usage being correlated with greater revenue performance people reminds me of a conversation we had in our newsroom recently we were talking about using AI in some way and it added to your said to me you know I've seen people who work with AI and people really don't want to work with AI and my observation is that people who work with AI wind up with much better jobs than the people who actually who actually refuse to work with AI and it shows up in our thrifescores which are our employee engagement scores again our employees who are using AI the most are also the happiest and if you look at kind of what are the tasks that they do during the day one of the things we're starting to see and we'll try to you know quantify more we're we're rigorously is that they have more varied kind of jobs and they're doing

18:11more interesting things. Alexis what do you so Katie's just described the situation at Microsoft and across McKinsey and your clients where are you finding organizations are at in terms of what you've laid out in your in your report like where things are at. Yeah I mean well I won't surprise you that Microsoft is on the front end of the innovation and the change curve I'd say you know I see companies everywhere on the spectrum from leaning all the way in and some of the very public examples that include some pretty public stumbles in some cases but very innovative experimenting thinking about rebuilding the entirety of the workforce the ways of working the delivery around an AI-based future and then I was with a CITRO last week I said well the first thing I need to do is convince the executive team on the business case to begin incorporating generative AI tools and I thought oh no we are we are not we're not even in the middle of the

19:13pack here we're you know we're behind and I think what's paralyzing for a lot of executives is there's so much ambiguity but this is moving so quickly that there's a natural instinct to say let me take a bit of a conservative and weight approach till there's more clarity so that I don't expend a lot of energy distract the organization and you know blow a lot of resources on something that may not work but the challenge underneath that is that really this is a change management game and it's a long game I mean we talk about that in the report and you need to be starting now doing all of the types of things that Katie's describing are just the way they think about deploying inside Microsoft which is you need to be thinking about building the muscle of experimentation of fluency of judgment around what works what doesn't work to keep moving the organization forward and becoming the type of company that has inside it the type of employees who are more AI forward leaning than back I couldn't agree more and I think in addition to

20:19the long game on change management from a people perspective there's a long game on change management from a data readiness data access kind of systems readiness perspective for big legacy companies and there's a long game on actually really understanding your work processes you know one of the things that is you know it's hard to tell AI to complete something if you can't actually articulate what that work process is and a lot of knowledge work is tacit and it's in people's heads and each of us see a little piece of it and we don't actually know what productivity or quality looks like and so there's actually a lot of work that needs to be done to make processes more explicit so that we can actually reinvent them with AI I think one of them more exciting applications of AI for me in some ways is is it's ability to make tacit knowledge explicit so you and I might be working you might be working on a project and there's lots of tacit knowledge that we couldn't actually easily share with each other whereas AI is able to to maybe help us actually do some of

21:20that Katie just mentioned or Alexis just mentioned fluency actually less just mentioned AI fluency is one of the things that organizations need and I knew I wanted to ask you about that in the report you talk about AI fluency being something that's in high demand what does it mean to be fluent in AI today and in the future like what is concretely what is what is that until I think it's a range of skills that we use to collaborate with AI so this is about using and sort of managing your work with AI so things like you know prompting right prompting is basically to have some notion of the question that you want answered to be able to frame that question in a way that AI gives you a reply usually it's not the perfect way that you framed it and you figure that out when you see what AI is telling you so the ability to iterate and improve and refine and sharpen your prompt to get what you want the sense of validation to kind of do the sense checks on the kinds of

22:26outputs that you get back from the AI tool and your ability really to refine that so I think all of these are now increasingly showing up as kind of skill tokens or words that people want right when they describe a financial analyst or when they describe an operations manager or when they describe a sales manager they're asking for the ability to work with some kind of AI to do these kinds of things yeah if it seems to be part of it it's just a willingness or like you're to experiment and to have you sit things that you described like being able to prompt and do things like that are things that it's almost a question whether someone has tried has been willing or interested or to try right I think it's mindset but it's there is a real skill set if you think about the skill of being able to articulate clearly in end state yeah to break work down into its sub components to actually think about what good looks like for each subcomponent and to delegate to then actually do quality control to communicate prioritize repriortize that's actually a set of

23:30management skills that most of us develop pretty late in our career right as we manage people and some people are better than that and others even who have management roles and now we're asking everybody in the workforce to be able to do that we did a workshop with my playbook we do these workshops with with teams to try to help them think through with different framings where the AI solution space would be that would really help them get to the next level and with one team the very first question I asked was okay pick one area where if you could really accelerate your performance it would really matter to your customers it would really drive your business results and that was that was really meant to be just the throwaway to get started so that we could use that as an example for the rest of the workshop most of the workshop was helping them think through what that really was and that's going to be the starting point so I do think there's a set of it's not just mindset to experiment it's also having perspective judgment this is why I think we're seeing that just as Anu described the best AI transformation success stories are always

24:32a collaboration between humans who have really great judgment and perspective and an understanding of what value is with AI rather than than not so as I'm listening to I'm hearing almost two sets of things so there's the management skills and Alexis I'd be curious to your reaction the management skills which is as you said to be able to plan and delegate to AI in some ways and evaluate the work and so forth there's also expertise and I think you know one of the questions that the initial research about the impact of AI and in a few of the initial studies were on call centers was that it supported helped the the least skilled workers make the most gains and actually the expert workers actually didn't see as significant gain but my read of the latest research is actually the the picture is more nuanced then there is a place for expertise be interested Alexis in your in your reaction to these two things like the management skills and

25:32the actual expertise in that yeah I do think there are different types of skills and we have a tendency to sort of smush together re-skilling up-skilling AI fluency which in the report actually breaks that down into technical and non-technical and the biggest growth is as Anu is describing in the non-technical like the average person changing their job to incorporate AI inside not the super coder the person who's going to create the agent or data environment that allows it to thrive and so there's a lot of different kind of lenses to it on the expertise point what I think AI now does incredibly well especially as the models advance is really point expertise and very specific domains that in the past you had sort of a scarcity issue in a lot of organizations around like we have one person who knows how to do X in this environment and now what you're seeing is in part because a lot of those places are places where you can really define or gather over time sort of

26:34the underlying knowledge base where the expertise is really a premium with the human is that judgment and how do I take sort of a situation and think about just as Katie was describing how you break it down into its pieces where the biggest value is going to be bring back an insight and look that at and say is that going to be material in changing you know a client or customer experience or not and those are very human things and so when you look at the skill curves that this research decomposes across every type of role what you see is there's some skills that we used to value immensely that are very point expertise driven or built over years of time I know how to use certain type of coding I have you know experience doing a lot of data mapping or working with the sales plan things like that that AI is basically automating away and then there are other skills that are incredibly human at the heart negotiation skills pattern recognition skills you know prompt

27:34and judgment skills that AI is really not very effective at or not really meant to point itself at and so I think what's going to be fascinating to watch is even if a much smaller number of jobs go away which is certainly my hope what you will see is just about everybody's job description is being rewritten in real time yeah one thing just to just to stay on this point for a second Alexis one thing that I've heard and is that organizations are finding there's this idea of AI workflow so basically you have people who just sort of get the AI to spit something out they don't think very critically about it and then they hand it over to their colleagues this could be a presentation or an email or something and it's for their colleagues actually like sort through and one thing that I've been hearing is that the value of expertise is that it allows you to to better refine the output of AI and that this idea that less skilled workers are upskilled reaches its limit because they're not able to distinguish between what's

28:38slop and what is actually quality product as easily as an expert that affair yeah I mean I was is funny I was having this conversation the other day with a CHRO who said well I'm trying to use AI to improve my talent attraction process and my job descriptions and the outreach to a broad set of folks but then everyone else is using AI to create a job resume that meets my job to say and I now feel like my AI and their AI are like battling it out and I'm definitely not getting a better answer as a result but internally you have sort of the same phenomenon right like I'll have AI do answer my questions and do my work and then I'll pass it to you and and eventually the buck stops with someone and a leader is left trying to make sense of it I do think there's increasing discipline and some of that is the hangover of you know early days of experimentation models and things that are not quite as strong but I also think it's putting a lot of pressure inside organizations to say how are we going to continue to build that type of judgment and skill

29:39because if that sits right now in senior leaders because they have 20 30 years of the rear view mirror to work off of but the next generation is like oh it's all gonna fly out of AI and I won't need to know that I mean that's where you get this tension where you say no no you still need a ton of human judgment we just have to rethink where the humans in the loop and how the human shows up and I think in a lot of organizations we'll have to rethink how do you build a healthy pyramid of employees yeah pipeline all the way up who can learn those skills but learn them in a different context yeah we should well let's come back to the early career workers in a in a minute and you we've talked about a bunch of we talked about AI fluency some of these management skills some of the expertise the report actually is quite detailed about the skills that will remain relevant and will be remain less relevant over time can you talk through examples in both of those categories yeah sure I think there is a real theme of if you think about this skill change or skill

30:43impact curves right that we draw out for different occupations and sectors there's this real theme about skills that occur on the right hand side of that which are more likely to kind of get automated away are the skills which are more kind of codified data and information driven where it's really about processing something that is frankly quite routine right and pattern based so for example in manufacturing something like scheduling production right like or even estimating how much you should produce of something right if you think about health care or hospitals it's stuff like checking you know coverage specifications and details that's actually quite it's it's just mired in information and data and the patterns are very easily visible and it's something that you know you you could easily see being being automated away but on the left hand side you get many more skills where they're fundamentally more ambiguous and they typically tend to be things

31:44that involve more social and emotional layers of judgment ambiguity and judgment so there you see things like you know frankly even coaching a coach might use AI as a tool but the the quality of the coaching and the conversation might might just be something that we still believe humans should do right so things like that relationship and trust building selling techniques stuff like again you know motivation and management of performance management dialogues not just the cranking of the numbers to know what to say but to actually creatively ideate how performance can be improved so those kinds of skills are those that we think they will change too but they won't be automated away they will be things that we hope will be enhanced or become super skills based on a lot of things that AI can do for us and in the report there are a bunch of skills including communication management operations problem solving leadership that you serve at these we see

32:49these these kinds of skills continue to be relevant I would also say they are what we call in the research I think they are transferable skills they are so foundational they occur in most roles and occupations and they are somewhere in the middle zone of our skill change so we see a lot of potential there in things like problem solving and communication which I think we've all experienced using AI that you can get better at it if you use AI but it's not necessarily something that you would completely you know delegate away and have no role to play in yeah I think a question just to zoom out for a second then I want to hit some other topics like the early career workers and others but I think a question that a lot of people have is like for young workers and I asked this partly because you're looking at the skills that transfer what did I study what kind of job should I steer myself at for you know it's a parallel question for middle and later career workers what should I try and upskill myself for what should I try

33:50is there you know I'm going to recommend that everyone reads your report and then the answer will be totally clear to them individually but how would you you know Alexis maybe do you have a I mean I have I have three teenagers so I'm asking myself this question a single day and I'm chuckling because honor and I talk quite a bit about the single biggest question we could ask we go I read your report it's so interesting what I'd love to talk about is what major should my child study I mean that's to be the next report what's the answer what would I say to that I think what I would say in this moment is the best thing every student can do right now is think about whatever they're studying how to incorporate AI into their experience because I think the biggest challenge frankly is for some of the early workers who were one step ahead of this AI revolution and coming out into the workforce but not yet far enough in to have a lot of that expertise we're talking about that helps create sort of a bit of the the judgment layer and managerial

34:55layer that assists you and you're kind of in a bit of a catch up moment but for those who are sort of on the front end of the journey I think it's thinking about how to do everything with it my biggest worry in education is that we'll create too much of like a walled garden and say don't use AI for anything except checking for plagiarism don't use AI as a study guy don't use it in the environment of learning in fact we're going back to blue books in the in the room right because we're trying to manage the risks associated with it but you know it's it's like suggesting you're not going to use a computer when five years from now you know every job that has a desk is going to have a computer attached to it and so I think that's the biggest piece that that I focus on is that you know the famous McKinsey case interview you're now asking people to do the case interviews with AI which means they have to come to the table knowing how to use modern tools in order to address the kind of I think that's going to be interesting and I think that's going to be much much more common some of the people are hiring we want them to do a full

35:56stack product build with AI that's that's the interview as opposed to you know demonstrating foundational you know skills etc and it isn't that we don't need foundational skills we do but I do think I very much agree with what you said which is we need people prepared to to be the reinventors of how things happen with AI and so we need people who are really comfortable experimenting leaning in okay I think there's there's I completely agree with what Katie just said and I think there's an attribute or a or a dimension to being an experimenter and a creator who tries new things that is also worth encouraging and learning outside of the in the context of AI so if there is an attribute I don't know if you can call it a skill or a college major or something like that but if there is an attribute that the education system should really focus on building now it is that it is this mindset of tinkering and let me try something and let me create something

37:00and kind of figure out how to do something better it's that kind of approach to problem solving that you really want and then of course you want kids to do that with AI and to come to the work force more ready to be able to apply those AI tools but at the heart of it this is not an era in which we will be valued for knowing something and then putting our heads down and kind of doing just that thing this is something that's going to ask us to keep questioning and trying to create and do something different and better I don't I don't feel like I have the answers this but I do I have heard some interesting answers to this sort of careers that are more protected and like one thing we're seeing probably is around sort of people care and health care and things like that in the latest job reports the it's very clear that health care jobs are the almost the sole engine of job creation while tech for example is showing much reduced demand or creation of jobs Kevin Ruse of The New York Times once you know I've read in his book he talks about how if you do a job

38:04that people will pay to watch you perform your job like a musical performer or a professional sports person or even a celebrity chef that job is is maybe safe because because there's a little niche if we all become a fellow players you know if you start young with your team your team you have hope here another way of thinking about it is think about areas where there's more demand if things if the product actually gets cheaper so among the things that AI does is it potentially reduces the cost of things so if you work in a area like one example I think of is like medical scans like medical scans are very expensive but if we could get five dollar medical scans like we probably get a lot more medical scans and I think that applies far more broadly than we think so a lot of the places that we're seeing inside Microsoft these case examples a lot of the value of AI speed and it turns out that when you do something well faster or with a different quality

39:07levels I described before the demand for it goes up so for example one of the groups that is really reinvented the way they work is this group that does complex prototyping engineering projects with our big customers and so they bring together data scientists designers software engineers and they've figured out how to really change the way they work together and independently in their areas with AI and now the kind of prototype that might have taken 12 weeks before takes a week and so that's a totally different value proposition to our customers so the demand for this capability is way up and I think we see that over and over again I was actually talking to somebody from one of the big four firms and he was talking about how this group in India does a certain kind of analysis for their clients and it used to take about a month to do and it was you know whatever now it takes just a couple hours and so somebody said oh so if you've been able to reduce the size of the

40:09India team dramatically to take advantage of this great cost savings he said no now our clients ask for daily instead of monthly because it's valuable every day and so I think we're going to see a lot of that kind of dynamic not just in the you know in the the medical scans and things that are obvious but other places that are less obvious yeah that makes a lot of sense so we encourage people to think about those areas where if things can be faster or cheaper the demand actually goes up yeah when I think about journalism I'm afraid that if we produce more articles cheaper the the equation not in our not in our favor there's a there's an essay right now by a tech CEO Matt Schumer that's sparking a lot of anxiety and reaction about the impact of AI on jobs and among the other things he sort of thesis is basically we see where coders are sort of on the front lines of the waves that's gonna that's gonna come through the tsunami that's gonna go through the jobs market

41:14and that what he said is basically AI is doing codershops and his his sort of thesis is that we all should be do things like get our financial house in order and yes and things like his his view is actually pretty alarming you kind of start hyperventilating when you when you read it and he said I know the next two to five years are going to be disorienting in ways most people aren't prepared for this is already happening in my world it's coming to yours Katie do you want to react and maybe I love other's reactions too like the his primary thesis is AI today is so much more powerful and more effective than AI even a couple months ago because of the most recent model releases and we are definitely seeing this and we're seeing it first in software and so people are you know it's blowing people's minds is what what can be done today different and so I talk about the

42:15case studies that we've done one of the questions I have is with the new capability of AI will some of the findings that I described change and how fast will they change and will they change so we take that seriously but I was talking to one of our engineering heads actually about this not about the essay but before that about the new models and he said just to be clear I want to create I have such a long backlog of things that I need to do for my customers I want to get through that faster and I want to create twice as many products and and releases but that are in a completely different level of quality that are you know where I'm using the speed and the capacity to do much more customer interaction and really understand what the customer needs are as opposed to doing 10 times more at the same level of quality that I have today and so again I think we're going to see the that even the new capability that AI brings to the table applied in a way that is aligned to where the greatest value is which may not be just taking human capacity it out it's probably going

43:19to be quality and speed as well and then the other thing I question and we'll see is I think that encoding which is only a part of the end-to-end product development process right I think we are seeing you know the kind of productivity that that Matt describes it is hard to apply this to finance and processes and payroll and sales and HR etc because again these are not these are complicated processes workflows that have lots of systems and data and different people doing different things without a lot of clarity is to what they are actually doing and what the end product is and how to redesign and so I think there's going to be a stickiness around some of these other enterprise processes that maybe you're not it's true for coaching. It's like friction the change doesn't happen as instantly as it might incoming which given that it's all digital etc. I think so and you know we're already have been seeing AI native companies redefine those processes in a completely new way AI first but that doesn't mean that you can just pick that up

44:22and apply it into other larger companies. Alexis how do you see this? I mean I would agree fully that the example that he uses is you know is quite compelling about what would take coding and weeks maybe months that AI doesn't just build an initial instance it tests itself it refines itself it as a judgment he talks about adding taste and like these very human attributions to create something even better than he thinks with extra manpower more time he would have even come up with and it's a very compelling thought but then I think about every enterprise I know and you know they're not producing an app a minute I mean if you gave him the chance that's not how they would create value in the organization and so I think to Katie's point it's like it's a whole system and if you start with where's the value creation actually happen I think there's some places where that concept is really exciting like drug development the idea that you could take

45:24you know we've often talked about financial inclusion and the idea that you could take that kind of expertise that's delivered to the wealthiest and apply that in a highly tailored way to the masses like there are these great examples where if you can get that right you should unlock value but it's not because you're going to take you know huge swaths of the people away because the agents will will run the universe I do think one of the concepts he gets dead right is if you believe this is more revolution in a permanence and I do that's here now to stay in terms of technology deployment how much time do you devote personally to getting up to speed and I'm amazed when I asked that question in board rooms and executive rooms with management layers folks who are meant to drive the value you know an hour a day most of them might be spending an hour a week you know you might think you should be spending three days a week on this right now but the

46:26the point being like if this is going to be that profound in whatever form it takes the gap right now between how not just average employees but even leaders are keeping up to speed with kind of what's going on just so they have a sense of the potential is widening by the moment yeah I saw a recent survey of CEOs global CEOs across all different industries and they have categorized which CEOs are most bullish on the value that they're getting and which ones you can see kind of the retraining investments that they're making and the technology investments and one of the key differentiators between those that are really leading and starting to really change our companies is how many hours a week they were personally spending to learn AI themselves yeah and in the match humor essay he says I know the people who come out of this best are the ones who start engaging now not with fear but with curiosity in a sense of urgency which is exactly what you're all saying and you I want to we you and I spoke a few years ago about you did a really interesting

47:32research about how labor market demographic factors meant that labor markets were tightening basically they're just there's less immigration in the US for example there's an aging population and the US like other companies could actually struggle to find workers AI was was a factor at the time but if you overlay what you know now about AI do you have the same analysis that labor markets going forward in a country like the US or in the US will be tight because of a insufficient supply of labor yeah I think that's a really interesting sort of confluence of these two trends and it's almost I think surrender pity right that we are on in the midst of this AI revolution at the same time that we are in the midst of a massive demographic reset around the advanced economies at least if not the whole world right yeah I think back in 24 when we we've seen what felt like peak levels of labor shortage and labor market tightness in the US and Europe a lot of that shortfall

48:39of labor or open vacancies which were being unmet in the job market were interestingly in work that was more physical in nature right so we come back to health care related occupations construction accommodation and food leisure and hospitality some types of transportation it wasn't really at that point that we were seeing shortages in all of this information based you know the information economy was not sort of tight at that point it had attracted a lot of workers historically like everyone's gone into knowledge work everyone's into finance or consulting or tech or whatever what have you right and then in manufacturing you weren't really seeing it because again in manufacturing because of the industrialized nature of manufacturing automation rates have you know were were already trending up and you weren't really growing by adding more bodies into manufacturing so it was these physical work and physical skill let's say intensive types of occupations and I think going forward I mean in the long long run our numbers on demographics

49:43will clearly say that if we don't get the productivity kicker of AI we are poised to anything from half to 1% a year of lower GDP per capita growth in most advanced because you have constraints in terms of the number of people right who can work and therefore it is imperative to lean in and actually get that productivity boost yeah but that said even if we do with what AI promises now the numbers tell us that it's possible right to get that productivity boost but I think there's this really interesting resetting of wages and returns to different kinds of skills in the labor market that we might see because the conventional knowledge based work is going to get reprised differently yeah because some parts of it are actually going to be commoditized right and some parts of it actually will get super skilled but the more physically intensive work depending on

50:43you know robotics has its own possibilities but also its own limitations because of decentralized and unpredictable work environments and things like that but you will see a repricing of skills and that's another thing that I think all our kids should keep in mind as they think about careers which jobs there's been in Alexis maybe so there's been this you know what we've seen over the last like several decades is there's this idea that you that in the US that students go to university they graduate to relatively high paying or stable white collar jobs and it feels like we're at a bit of a sort of cracking point where graduates of the last few years of universities are facing a much tougher job environment and the role of AI I think researchers Eric Brinielson and others there's feel like AI is actually playing a direct role in the reduction in hiring of younger workers I'm curious like how you know what your what your own practice is and what you're saying across your clients is AI reducing the hiring of younger workers and is

51:49that you know is that a short-term thing or are we how will that play out yeah it's it's really fascinating right now because I hear folks saying like it's going to be a pyramid but it's going to be a flatter pyramid than other people know it's going to be more like a rectangle and then yeah we're a diamond or like what was the is the wine glass that really that really terrified me I thought oh it doesn't feel like that's going to work very well in a young economy but I think the jury is truly fully out on that and folks are the movement that we see as folks are all over the place and I do think there's also a question how much will this vary by industry because jobs and the makeup of jobs the amount that robotics or AI impacts you know the research from Anu's team it it actually looks quite different even today's picture across different industries and sectors depending on the composition of who you have in the workforce but all of that said I think you know that question is very acute and you do see a tougher job market being described certainly by

52:54very recent intrinsic to the job market and college graduates in particular some of the headlines that we see about all the jobs that have been taken out and are and all the restructuring underway in the last 12 months that's getting sort of the banner of AI is enabling us to do this I will tell you on the inside is just classic belt tightening yeah it's mostly the hangover from post covid growth that was either unbridled or sort of mismatched to long term the AI is cover cover and it gives you a great bond right I mean analyst love hearing that you get a little bit of credit for being on the forward end but really I'll tell you from the inside it's a lot of cleanup in most organizations more than anything but the implication is there are a number of folks who've really slowed down or even stopped entirely right now they're hiring I think that's a huge mistake because first of all what do you want in 5 10 years a pyramid of very expensive soon to be retiring individuals and no one you know fresh and new in the job market but also if you think

53:56about just where innovation comes from how you need a whole apprentice apprenticeship pipe a healthy company has you know ventilation if you will or flow in it in terms of the workforce and so you know my my belief is what we're going to start to see is that rebalances a bit as what becomes more and more of a premium is some of that general AI fluency that we already see playing out in job specs and while that shouldn't have age as a factor usually technology adoption moves more quickly on the younger end of the we're seeing exactly that and actually I was with a whole set of CHROs and we had this whole discussion and nobody in the room was going to reduce you know their no wine glasses no wine glasses no diamonds well there's one guy who said I think we're going to be a diamond then we're like that's ridiculous we're staying in a pyramid because of and and you know we were just at a conference actually the charter conference where Nicole Amaro from who's the CHRO of IBM really said look we're tripling down on early career

55:02hires and and she made the same case you did which is if you want to have a pipeline of the people who have the knowledge and expertise and experience to to lead in the future and be the experts of the future you need to hire them I think we're seeing two other things one you mentioned which is hearing up early career with later stage is a really wonderful way to get the best of both worlds so we're seeing lots of companies and we're doing some of this ourselves doing this reverse apprenticeship type of things because it's the young people come in who are native AI folks and reverse mentoring yeah reverse mentoring who are saying wait why are we doing it this way that's crazy we're going to change the way we work I've been playing with AI let's do it this way but they need the the the wisdom the judgment the taste of the more senior folks we have a new program called praise is one of our engineering teams which is a preceptorship which is like this putting together senior and junior people together so in my mind what we really need to do is redesign early career roles and redesign early career learning and development to try to accelerate

56:08all of these management skills and the judgment that we've all been talking about in early career but we need those early careers to also be challenging and bringing real energy and ideas to reinvent the way things happen for their colleagues we've a bunch of great questions here so let's spend a few minutes going through them and the first one is it feels like a lot of productivity studies this far look at individual role augmentation with AI what is your research showing about teams that use AI what are best practice I love this question because I keep telling people I everybody has a review of the year of 2026 is going to be the year of and I think it is the year of shifting from a real focus on individual adoption individual fluency individual productivity to team and system change and so a lot of what we're trying to do is helping customers but also our own teams really think about how do you work together to change the way you work as a group whether it's looking at a whole persona so everybody adopts best practice at this you know in the

57:10best way as opposed to lots of individual variation or whether it's a whole end-to-end workflow changing with all the people who are part of it one of our best programs we call camp air which is a team-based learning program where we bring entire intact teams of software developers so includes the designers the product managers and the coders and they go through a process together to get immersed in the latest and greatest AI technology and then to change the way they work together as a team and then to actually apply those learnings to the actual work that they're doing in the flow of their work and that immersive team-based approach is so much more effective than individuals going and trying to learn things so I very much agree with the point that we now need to start focusing on teams great we're focusing on teams here's a question from the audience which is is there a concern about processed knowledge decay if a single person can do the work of 10 what

58:12will the future of training look like and menial tasks are generally how new workers learn the systems of operation is the future most workers doing high-level work and without these lower level tasks how will they build expertise to an Alexis here yeah it's a it's a really interesting thought process to take forward and part of that is sort of the cognitive load or experience of the jobs of the future there's a great hbr article just the other day on this we've done work in our brain science health lab on this but it's like we say we don't like the menial stuff but in addition to training us on how to get better at the you know the corner cases the judgment layer above that it's also actually pretty easy to do repetitive menial work in a knowledge worker space and so there's a sort of cognitive relaxation that happens in your day even if you're like oh now it's a time when I have to do that thing it's super tedious yeah and if you just remove all of that

59:12and all you leave people with is like the really intense thinking stuff it's really exciting it also can really to the extreme take some jobs that were never meant to be I mean take the call center worker who only gets escalations of the worst type you know that is a tough job all day long and so I do think there's some questions underneath that in most organizations the best learning is on the job in the moment experience and so I believe what we'll start to see is as these processes are reimagined end to end just like Katie was describing the humans don't go away they're rolling the loop of the agent experience looks different the fluency they need looks different there will still be in the healthiest of companies a front and pipeline of people who do that work they will just engage in that work they're kind of learning journey will start in a different place and so we see that already with our analysts you know the work that they used to do we now have

1:00:13AI augmentation to enable them to do they still have a skill set of what they need to do but instead of spending hours toiling over the same set of analytics that comes to them and they begin their process in a different place and I think that's fundamentally what we'll start to see and maybe this question might be for you this this technological delta feels different than others in both time horizon and the disproportionate number of people that will be replaced for example when phone booth tellers were replaced by technology it took place over 15 years giving people a wider time horizon unless people fewer people at one time to reskill what is the best slash worst case scenario when it comes to reskilling and then it's basically will the pace the rate slash volume of AI replacing roles feels like it could result in a large percent of employment if we're not careful about that transition phase yeah I mean just harking back to the matchuma article we were talking about I think that does underscore this what is different about this technological

1:01:14revolution in that apart from everything else that we've talked about it is also a little recursive in the sense AI is making AI better and you're able to focus all of the productivity speed efficiency quality learning that AI brings to the table to make AI itself better right so that's already propelling momentum with and sort of exponential rates of change and things like that but I wouldn't I wouldn't minimize the adoption friction part of the story and in some ways the the pace of change the rate of change of the underlying capabilities also complicate adoption because you know how do you actually know kind of what version or what generation of you know your models you're really running with or embedding or wiring and if that's going to change what happens right so adoption is complicated for a number of reasons there are economic cost benefits

1:02:16of you know is it really more valuable to kind of bring in technology here and then there's this whole change management and just the friction around all of that so we think that relative to let's say theoretical potential of 50 or 60 percent of work being automated we think it's reasonable to expect in a five-year time frame that you will see those technology impacts in something of the order of magnitude of 25 to 30 percent right just a poor number there but it tells you that not all of those possibilities will just materialize overnight and we do have some time to kind of think through and orchestrate the change and the real let's say the big question about how the labor market finally will adjust does depend on our openness and speed of adjusting in terms of taking those productivity benefits and plowing that back into more innovation more new things that

1:03:18we can do the economy and the economic structure has to enable that higher rate of innovation building new businesses meeting new kinds of needs and allow for that I think economies that have kind of regulatory or or other kinds of setups that facilitate this will end up with better labor market outcomes eventually as opposed to ones that kind of don't allow the innovation and the technology different possible scenarios we just you know the Verizon CEO gentlemen to say he thinks most likely scenarios 15 to 20 percent general unemployment for precisely this trend if we don't manage that's that's not a consensus view and I know KDE you you don't agree with that and I don't think it is actually consensus view but we're in a moment of real transition and question so we're just about out of time I would love just like one last thought from each of you on on like I guess what you're your advice or you're sort of clear a

1:04:19certainty about kind of where we're headed is Katie do you want to go I'm I'll just reinforce this notion of learning agility curiosity experimentation that's more comfortable for some people than others but it really is going to be ticket to play and you know the good news is that we who do the knowledge work actually have to redesign the knowledge work because we're the only ones who know how it works it's not written down in some guide books somewhere so there's a real opportunity I think for for all of us to lean forward into this new technology and reinvent our own jobs in a way that will be more satisfying and will then give us the skills we need to continue to be relevant great thank you Lexus last thought I might add to that I think this is a journey that we're going to be on for decades if not infinitely and so the mindset in addition to learning is sort of a continuous motion that we need to be in and get really comfortable with and a lot of organizations and a lot of individuals are sort of wired for stability I think we've got to get ourselves sort of wired

1:05:22for perpetual change great and you I think just echoing that I think the AI we're seeing is the worst version of the AI that we will see in our lifetimes it's going to keep changing it's both but I think the bottom line is there's it's a call to action right there is a sense of real urgency about understanding and what you can do with it and moving forward and one of the takeaways I have actually hearing about your research is the demand for non-technical AI skills so these are people who are not sort of deep in the AI coding but actually are in business functions maybe or other areas where they have AI fluency and interest so we could keep keep talking but thank you for this conversation Alexis Crickrowicz and who Matt Kavar and Katie George from Microsoft and Microsoft and McKinsey I should tell the folks here that there are free copies of a century of plenty available just outside this room and you can also download it using the QR

1:06:22code on the side screens for more information about membership or upcoming Commonwealth Club programs please visit Commonwealth Club dot org I'm Kevin Delaney I'm the editor and chief of the San Francisco standards we're at sfstandard.com be very happy if you read our articles thank you for joining us thank you you've been listening to the Commonwealth Club of California here thousands of our podcasts on Apple podcasts Google Play and Stitcher if you like what you've heard please consider supporting our work help us bring 500 programs a year to listeners like you go to Commonwealth Club dot org slash donate think your way around the world with our travel program to exciting domestic and international destinations when you're in the Bay Area please join us for live events at our home on the waterfront thank you for listening and thank you for your support