AI Isn’t Replacing Jobs — It’s Rewriting Them

This transcript has been generated by AI and may contain slight discrepancies from the audio or video recording.

 Ad: [00:00:00] Everyone's panicking about AI stealing jobs, but smart companies are thinking bigger. What if AI helped your people crush it? Lattice is where people in AI team up to help employees grow. Managers lead and companies win. No regrets. Check out lattice.com/no regrets. That's latt IC e.com/no regrets.

Anne: Welcome everyone. Today we're diving into a major topic shaping the future of work automation, generative ai, and the risk of job displacement across US employment and HR. Now as headlines alternate between alarmist predictions and promises of productivity. It's easy to lose sight of what's actually happening on the ground.

This discussion will unpack key findings based on three SHRM research publications to clarify which jobs are truly at risk. YHR is in [00:01:00] the spotlight and how worker expectations around AI are evolving Now. Joining me today is Justin Ladner SHRM's labor eco senior labor economist, who will explore the data behind automation trends, generative AI adoption, and some of the barriers to displacement that often get overlooked in mainstream coverage.

Justin, welcome.

Justin: Thank you, and it's a pleasure to join you today.

Anne: Pleasure to have you. Now, before we get started, a quick note for our audience. We'll highlight key report takeaways throughout today's discussion, but be sure to check out links to the full research we've provided as part of this broadcast.

All right, Justin. Ready?

Justin: Oh,

Anne: ready to go? Yeah. Alright. So can you briefly, um, explain how the three reports that we just mentioned that we'll be discussing today fits with SHR m's, broader research around technology and ai, and what key areas do they address?

Justin: So, SHRM has done a lot of research, especially in the last few years, um, relating to.

HR technology, um, and, and, and [00:02:00] in a kind of nexus with, with AI and automation. Um, so this particular line of research, um, fits into that nicely. We wanted to, we had done a lot of work, done a lot of survey research that talked about, um, how people felt about AI in the workplace, um, their kind of attitudes and beliefs about it, how they might use it.

Mm-hmm. Um, this particular line of research, we wanted to fit in something that's specifically dealt with. Uh, the extent to which AI and automation, uh, might, uh, pose a risk for displacement. So that was the kind of central idea of how we wanted to, uh, expand on the existing research that Sherman had already done.

Anne: Right. So we, we've mentioned higher automation, displacement risk, and non-technical barriers. Mm-hmm. Can you kind of walk us through the definition of those terms?

Justin: Yeah. So back in, uh, this would be late 2024, we had been seeing all these headlines, um, to give a little motivation for mm-hmm. For how we kind of got to this place.

Um, either, you know, I'm sure people in the audience too have seen these headlines that're really polarizing. There's some headlines are saying, you know, everybody's job is gonna [00:03:00] be replaced by a robot. Other headlines are saying this is a fad, it's all gonna blow over. Mm-hmm. That's all a big bubble. Um, and we kind of wanted to create a source of ground truth.

Mm-hmm. And the existing research is, a lot of research on this topic goes back, um, quite a long way actually. Um, but it's, there's a lack of consensus. And we were, we had reviewed a lot of that work and we felt that we could make some improvements. Um, and, and these definitions kind of illustrate how we felt we could improve things.

So we wanted to create a system where we tried to identify an occupation or, or individual job as having a high, um, exposure to displacement risk. Mm-hmm. If it met a certain set of conditions. Mm-hmm. So there's two main conditions. One of the conditions is a high level of automation. Mm-hmm. And we define that, uh, as an occupation or a job.

Where at least 50% of tasks are currently automated. Um, and then the other condition is the absence of non-technical barriers. And a non-technical barrier in this context is any kind of obstacle to displacement that doesn't have anything to do with the technology. So imagine that, um, in, in theory, the technology exists to displace a [00:04:00] job, but there's some non-technical reason.

That, um, that that job still might be protected even if the technology exists. Good examples would be client preferences for human interaction or just to have a human in a role. Uh, I very often use the example of an airline pilot, so technically speaking, uh, air travel, certainly in a modern commercial airliner is largely automated.

But would you get on a plane if there wasn't an actual human pilot aboard? Um, another example would be legal or regulatory barriers. So there are many cases in which, um, legally, uh, or to follow regulations, um, automation is impossible. A full automation is impossible. Mm-hmm. You have to have a human present.

So those are a couple of examples. Another one might just be cost effectiveness. So there's plenty of context where in theory you could automate something. But it actually is just cheaper to use human labor.

Anne: I always love the airline pilot example you guys provide, so yeah,

Justin: it illustrates it pretty vividly, I

Anne: think.

Yeah, it really does. Yeah. So let's dive into the first of our three SHM reports. This one is called Automation, generative ai, and Job Displacement Risk in US Employment. How widespread is automation really across the [00:05:00] US workforce and how many jobs are at least 50% automated?

Justin: Yeah. So that first critical condition was that meeting that barrier of, of being highly automated as a job.

Mm-hmm. So what we did in terms of approach to this research is we, we have a framework involved, a very large scale survey. Um, to basically estimate for every individual occupation. Um, in this BLS data set, we estimated, um, the, these key automation and, and AI sort of exposure, uh, factors, including the share of employment in every occupation that we estimate to be at least 50% automated.

And then we can aggregate up to talk about any groups that we want to. So for example, we can aggregate across all occupations and talk about total wage and salary employment in the United States. And to get back to the original answer to the question, right, um, we estimate that about 15.1% of, of total wage and salary employment in the us.

Is at least 50% automated. Mm-hmm. And as of the time that we were using that data, this was BLS data from May, 2024. Um, the, uh, that number 15 1% equates to about 23.2 million jobs.

Anne: [00:06:00] I was about to say the 15.1% comes off a little small, but then you put it into the real numbers and Yeah. It's millions of jobs.

Justin: Yeah. I, another, another way to think about it. It's about one in, one in six, one in seven, uh, workers, one. Yeah.

Anne: So does the degree of automation in jobs differ by occupational groups? And if so, what patterns or variations have you observed in this research report?

Justin: Absolutely. It differs dramatically by occupational groups, um, and in, in a lot of ways that you would, you would very much expect So on the top end.

By far the most, uh, the occupational group that is most likely to have very highly automated jobs is the, what's called the computer and mathematical group. Mm-hmm. And that includes a pretty diverse set of occupations, but think about it as being computer programming, software developing. Mm-hmm. A lot of it occupations, a lot of kind of database engineers, that kind of thing.

Um, these are occupations that interestingly, historically have had really limited exposure to automation, um, that they're just not the types of jobs that you would think about as being classically associated with automation. Um, but, but recent sort of trends in [00:07:00] technology, particularly generative AI tools and other types of AI tools have really changed the game in terms of what can be done, um, that previously was required, you know, human intervention intervention.

So a great example would be computer programming. Mm-hmm. Historically, you'd had to have a person write a computer program. Right now we have really sophisticated AI tools that can actually do quite a good job at writing computer programs. So that group stands out as being a really high group groups that stand out on the low end.

Again, it's pretty intuitive. Mm-hmm. Um, something like food service. Mm-hmm. Um, it's just, it's just not a, it's not a particular occupational group where automation comes in very often. Uh, labor's very inexpensive in that group, so it oftentimes doesn't make sense from a cost point of view. Mm-hmm. Um, another group set of groups that stick out on the bottom are healthcare type workers.

Yeah. And education type workers. Again, for some similar reasons, but those groups also have this feature of being kind of inherently human. There's a very. Strong desire, um, to have a human play. Uh, at least, you know, a very central role in that. So there's relatively low levels of automation,

Anne: I think we can all understand that well too.

Yeah. And, uh, you [00:08:00] kind of led into my next question about, uh, generative ar ai or gen AI for short, as we call it, is becoming more commonly used in the workplace. You know, as technology advances, what percentage of US jobs are now using specifically Gen ai and. How many jobs does this represent?

Justin: Yeah, so the, in the report itself, we don't talk about the fraction or the share that actually used some positive amount of generative ai.

That number is pretty high. It's about 60%. Mm-hmm. I have to go back and look at our, the actual underlying data to see the numbers, but basically generative ai, just, just in terms of its presence in, in US employment, um, given that it's so new is remarkable that, you know, over half we would say of jobs have at least some usage under bay.

Now the share that use at least. That complete at least 50% of their tasks via generative AI is much, much smaller as you'd expect. It's about 7.8%. That equates about 12 million jobs. Okay. Um, so, but again, thinking back and thinking that chat GPT just came online, this is less than four years ago. Um, it's a pretty dramatic shift in a short space of time.

So, [00:09:00] um, generative ai. Is certainly pretty rapid

Anne: chat. GPT feels like it came out much longer ago. Yeah. But yes, that's a good note. Uh, so, uh, similar to the automation question, are there notable differences in, uh, gen AI adoption across occupational groups? And really how does the share of jobs using Gen AI vary by sector?

Justin: Yeah. So, uh, again, with occupational groups, um, the ones that really stick out computer mathematical occupations, again, way ahead of everybody else in terms of using generative AI tools, not super surprising. That's mm-hmm. The sort of, the, the types of workers that are in that group are also the types of workers that created these tools in the first place in many cases.

Mm-hmm. So there's, there's a wide acceptance of the use of those tools. There's also. A lot of the tools are designed to be used in that context, so it makes sense that, that, uh, that group would be particularly kind of, um, affected by, uh, generative AI tools. Other groups that stick out are a lot of, um, kind of professional, technical, scientific type occupations.

Mm-hmm. Um, a lot of sort of business and financial occupations. So HR, and we'll talk about this [00:10:00] more. Yes. Um, is, is is a group where gender value tools are becoming much more common. A lot of, um, a lot of kind of business operations type occupations too. So even, uh, financial analysis. Type occupations or even accounting type occupations.

Mm-hmm. Generative AI tools have become, uh, pretty effective at handling a lot of those tasking. So, so those occupations also have seen an impact and, and a greater uptake, um, in terms of generative ai.

Ad: Doing great. Anne, let's take a quick break. Oh, hey, I didn't see you there. Well, while I have you here, if you're enjoying this episode, imagine sitting in as part of the live studio audience for an upcoming recording.

By the way, hi, I'm Melody, the producer of SHRM's All Things Work podcast. Our team will be recording live at SHRM Talent 2026 in Dallas, Texas, from April 19th to the 22nd. And you're invited to be part of the action. Join us for a front row seat and experience podcasting behind the scenes. For more details, head to SHRM dot org slash talent 26 podcast.

View the full event schedule under program and search podcast. Hope to see you there. [00:11:00] Alright, let's get back to the show and back to you.

Anne: The path to full displacement is a little more complicated than, you know, we might think it's not, it's not so black and white. It's gonna go to like immediate displacement.

Uh, because you mentioned those non-technical barriers that exist. Uh, so for jobs with those non-technical barriers, like the client preferences, the regulations, yeah. What are the most frequent obstacles to automation and how significant are client preferences in this context?

Justin: Uh, they're very significant.

Mm-hmm. So, so in terms of non-technical barriers. Um, we, there's sort of a two stage way that we approach this. First off, um, how likely is it that a job in a particular occupation is gonna have a non-technical barrier? And second, um, if, if a non-technical barrier is present, what are the types of non-technical barriers that are most common?

Outro: Mm-hmm.

Justin: So overall across all US wage and salary employment, uh, we estimate that about 63.3% of employment has at least one non-technical barrier. And if you dig into that share of employment, that that set of jobs that have non-technical barriers. [00:12:00] The vast, the, the, the most likely, the most likely non-technical barrier to exist, um, is, uh, gonna be client preferences.

Um, and with about almost three quarters of, uh, of people that report at least one non-technical barrier, um, that, that barrier is, or one of those barriers is, is client preferences. And again, it makes a lot of sense. Um, it does, yeah. The types of occupations, again, HR is actually a good, a good example.

Mm-hmm. Um, having a human in that role, uh, is, is just heavily valued, uh, healthcare education. Uh, you know, you can kind of, you can think about all the intuitive occupations where you'd want to have a person mm-hmm. Um, present, uh, that, that, uh, that comes up quite a bit. Um, and, and those are where those non-technical bears are really important.

Um, but there are other non-technical bears that come up pretty frequently too. Mm-hmm. So legal and regulatory rank second for us. About 40, about 41%, 41.5%, if I'm remembering correctly mm-hmm. Of, of employment or of of jobs that we estimate have at least one barrier. Um, have a legal or regulatory issue. Uh, and that could be something, again, HR is actually a great example.

Yeah. Um, if you're providing somebody some advice or some kind of [00:13:00] guidance that's gonna have a legal ramification. Um, it's probably really important to have a human in that role. Mm-hmm. And if you don't have a human in that role, you might be violating a regulation or violating a law. Again, the airline pilot rule, um, there's a lot of states, a lot of localities where something like a, like a driverless vehicle, um, wouldn't currently be legal or it would violate some regulation.

Mm-hmm. Um, so that would be another example. So that comes up quite a bit. And finally, cost effectiveness, um, is sort of the third one that we, that, that we looked at in, in sort of a specific sense. Mm-hmm. That comes up quite a bit in occupations where. Labor itself is just very inexpensive. Um, so it might be that you have a, a job, a a, a really good example might be a cashier, for example.

Mm-hmm. Um, where we actually know that tech technology exists. We actually, we know it's used in plenty of context, right. We've all been to stores where the only way to check out. Um, is through an automated process, a self checkout kiosk. Um, but we still have plenty of cashiers. There's about 3 million cashiers in the United States.

Mm-hmm. And one reason for that is that in a lot of contexts, it just doesn't make financial sense for a firm to invest in automation. [00:14:00] Mm-hmm. So if I go into a mom and pop shop Exactly, it's very unlikely that I'm going to be checking out at a kiosk because that kind of investment just doesn't make sense.

It's much cheaper to actually hire human labor or just the owner themselves to, to just deal with that. So, so there's plenty of context where even if the technology exists. Um, it just doesn't, from a cost point of view make sense for the firm.

Anne: That makes a lot of sense. And now kind of going to the other end of the spectrum here, for the jobs that are highly automated already, maybe at least 50% or more, and lack those non-technical barriers, what share of view as employment falls into that category and what implications does this have for displacement risk overall?

Justin: Yeah. So that if that kind of gets to our second condition mm-hmm. For identifying a high risk. Sort of, uh, job is you are both highly automated. At least 50% of your tasks are automated and you don't have those non-technical barriers. So, um, just to kind of bring the audience back to that original share, the share of employment that we estimate is at least 50% automated is 15.1%.

But if we add in that [00:15:00] additional condition of having no non-technical barriers. And the share of employment that meets both of those conditions falls all the way to 6%. Mm-hmm. So it's a much, much smaller share. 6%. It's about, um, if I'm remembering correctly, about 9.2, 9.3 million jobs. Yeah. So again, that's not nothing, that's still 9 million jobs that we estimate have this kind of high exposure to displacement risk.

But it is, if you think about total US employment, it is a small share of overp employment. So it's something like one in 17, one in 18 workers

Anne: puts us into a reality check a little.

Justin: Yeah. So it's significant. I mean, if, you know, if. You're one of those 9 million jobs if you're, you occupy one of those 9 million jobs.

That's certainly, um, something to, to be concerned about. Um, but it is still the case that we estimate that it's mm-hmm. The vast majority of US employment, at least currently in the near future, um, doesn't have this sort of imminent threat of displacement.

Anne: Makes sense. Okay. So we've been teasing head to this.

Let's finally turn onto this second. Second, um, of SHRM's. Three reports. Today is called Automation Generative ai, and Job just. Placement in HR employment, we teased ahead, so let's finally dive [00:16:00] in. It's essentially that deeper dive into all the US occupational data you collected focused on just the HR roles.

So can you walk us through why you decided to focus specifically on HR roles for this separate report? I know you touched on a little bit, but let's elaborate.

Justin: Yeah. So when we. The, when we kind of constructed, we always thought about this as a line of research that was gonna produce multiple reports over many years.

And actually we're, we're gonna be extending that this year. We're gonna do a new survey, we're gonna have a whole lot more research, uh, coming. Mm-hmm. So that you have all that to look forward to. But when we first constructed the original work, the, the, the broader report that we just talked about mm-hmm.

Um. We intentionally had this structure where we were estimating these values for every individual occupation. And the reason for that is that we could then aggregate to talk about any group that we wanted to. So if we wanted to focus on a particular set of occupations, we could do that. And in this case, um, we thought, okay, we've done all this work.

We know that this is a big issue in HR and a big concern for a lot of the, um, people that we talk to when we go to conferences. Um, it comes up quite a bit. Um, we know that, uh, generative [00:17:00] AI tools especially are really expanding in HR. So we felt we should, we can just very easily take the analysis. That we've already done, and we can focus on HR occupations specifically.

So in this occupational coding structure that we focus on, there are about 10, um, occupations that are specifically HR Examples include, um, hi basically, uh, human resources Managers is one of the occupations, compensation and benefits managers, training developments managers. Um, we have kind of HR professional specialist type roles that cover similar areas.

And then we have a few sort of, um. Lower level roles like HR assistants and payroll and timekeeping clerks. Mm-hmm. Mm-hmm. So we have a set of 10 occupations that are very specifically related to an HR function. And so we thought about those groups. We basically examined what we had found for that group as a whole and for the individual occupations in that group.

Outro: Some days it feels like you're just running to kipa. With SHRM, you set the pace and you stay ahead. SHRM membership now gives you access to business intelligence, advanced benchmarking, and in the moment, support from HR experts rely on SHRM because if it's a work thing, it's a [00:18:00] SHRM thing. It's.

Anne: You touched on how generative AI is kind of the use of it is growing in the world of HR.

Yeah. How prevalent is automation in HR roles now and how many HR jobs are already at least 50% or half automated?

Justin: Yeah, so we estimate that about 19.1%. So nearly one in five HR, uh, roles currently are at least 50% automated. Mm-hmm. Where the, the tasks that are done have an automation component in them at least 50% of the time.

Um, and now that, so comparing that to overall US employment. The overall value for US employment was 15.1%. So 19 1% suggests that HR is a little bit more exposed to automation than the kind of average American job.

Anne: That's amazing. So how much are HR workers really using generative a i to complete their daily tasks?

Do you, you kind of touched on this a little, but I would love to elaborate. Do you we see a higher number of HR jobs, at least 50% automated by this type of technology?

Justin: Uh, yes. So, uh, Geneva, even more so, we actually. See a bigger, mm-hmm. A bigger jump. Um, in percentage terms, um, relative to, um, just automation in [00:19:00] general.

So overall, US employment, we estimated that about 7.8% of jobs had meet that threshold of being at least 50% done using generative AI tools. Within HR, it's 11.9%, so it's again about. A 50% jump in terms of the, the size of that share? Um, so it's a, it's, it's pretty significant about one in, one in nine, um, one in almost one in eight, uh, HR, uh, jobs we estimate is at least 50% done using generative AI tools.

Anne: Oh, wow. So bringing back those non-technical barriers. 'cause it's such a. Big part of this discussion. We talked about it, how it looks in US employment, uh, that research report, but what do these barriers really look like in the HR field, and how significant are they compared to the previous research report we just discussed?

Justin: Yeah, so they're even, as you'd expect, and as I alluded to, they're even more prevalent, um, in HR. Yes. Um, overall, we estimate that about 64.4% of HR jobs have at least one non-technical barrier. Mm-hmm. And of those jobs that have non-technical barriers. Um, again, very much as you'd expect, um, [00:20:00] the, the kind of barriers that stand out are client preferences and, um, uh, legal and regulatory barriers.

Mm-hmm. So it's, um, the number for client preferences is something in the high seventies, over three quarters of HR employment mm-hmm. That has an non-technical barrier. Um, over three quarters of them have a, a client preferences barrier. Um, and then over 50%, slightly over 50% have a legal regulatory barrier.

Um, so those, again, for a lot of HR roles, those end up playing a really, um, a really big. Role in, in reducing the amount of HR employment that we actually mm-hmm. Assess as having a high displacement risk.

Anne: So you say those are kind of the most common obstacles in preventing the full displacement?

Justin: Yes. So, so if a non-technical barrier is present, those two things by far are the most common.

Anne: Okay. So we, we, um, when we talk about this research, I know you and your team talk about how it's more of a job transformation.

Justin: Yes.

Anne: And, uh, so, but the reality of the research is that. Some jobs are still at risk of full displacement. What proportion of the HR workforce really falls into this [00:21:00] category? Can we explore that a little bit?

Justin: Yeah. So, uh, the overall share of HR employment, again, we go back to those ideas of which, what share of employment meets both of those conditions to mm-hmm. To, uh, to kind of fall into that high risk of displacement category. For HR, it's about 9.3% of employment. Mm-hmm. And again, compared to, um, the overall number that we had.

Said that about 6% of all US employment faces, um, that meets both of those conditions and is therefore highly exposed to automation displacement. In HR, it's about 9.3%, so it's a bigger share. Um, and it's, you know, about one in 11 HR workers, so it's a significant risk. Um, but again, the vast majority of HR workers, you know, if we put this in the context, over 90% of HR workers, we estimate don't meet both of those conditions.

Okay. And therefore have at least some installation from displacement in the immediate future.

Anne: So what, what does that speak to for the field overall, would you say?

Justin: Well, I think you brought up the point of transformation. Mm-hmm. And I think that is really the most salient point. So the, the fact of the matter is, is that even if.

An HR job right now isn't dominated by [00:22:00] using automated tools or generat AI tools. Mm-hmm. Those tools, uh, always pretty much always know in HR at least have some relationship to tasking they're related, that they are involved in at least some tasking, and the idea would be that that will certainly grow in the future.

Mm-hmm. So if we think about the types of tools they get used. In HR most often. Right. Um, they're increasingly being, they have some kind of AI component. So the idea that you would continue in the HR profession and not come across those tools mm-hmm. Um, seems increasingly unlikely. So, so I, I think that the broad story for overall US employment and but, and certainly, um, a subset like HR in particular, is that, uh, integrating these kinds of tools, these kinds of technologies into work is going to be a, a reality that either is already here for many people or will very, very shortly arrive.

Um, so that transformation story, I think is, is the dominant story that comes out of this, um, this research. It's not that certainly displacement will happen, um, but it's more that people will transform the way in which they work.

Anne: Justin, thank you so much [00:23:00] for sharing your insights and your thoughts today.

We really appreciate walking us through, uh, this research and for our audience tuning in. Remember links. To the research that our conversation was based on. Today is included as part of this broadcast, so make sure to check all of it out. Thanks for tuning in, and we'll catch you next time.

Outro: If you enjoyed our conversation today, be sure to subscribe wherever you enjoy your podcast, so you never

Anne: miss a new episode.

Also, did you know all Things work is more than a podcast? That's right. We're also a weekly newsletter that includes in-depth articles and the latest research from SHRM to keep you in the know, just head to SHRM dot org slash all things work to sign up. Plus follow SH RM on social media to view the latest clips and join the conversation on game changing topics that are redefining the world of work.

Ad: Lattice is the HR platform where people plus AI succeed together to help your workforce crush it. No regrets. Visit lattice.com/no [00:24:00] regrets

SHRM