The Impact of AI on Labor Productivity and Growth
11 Feb 2026 · 25 min
In this episode of the Metrics Brothers podcast, hosts Ray Reich (Benchmarkit) and Dave Kellogg (Balderton Capital, independent consultant, author) explore labor productivity—a key but often misunderstood economic metric—and how artificial intelligence may reshape it. They begin by explaining the U.S. Bureau of Labor Statistics’ definition of labor productivity (real output per hour worked in the non‑farm sector) and highlight what the measure excludes (farm work, government, non‑human labor, overtime for exempt employees). They discuss historical trends, noting that productivity per hour rose from about $60 in 1980 to $104 today, roughly a 70 % increase, and compare the U.S. to other OECD nations using purchasing‑power‑parity data, with Ireland leading. The hosts examine productivity growth rates, showing a slowdown from an average 1.8 % annual increase (1950‑2000) to about 1.2 % (2000‑2020), alongside a decline in labor‑force growth. They reflect on why recent technology waves—SaaS, cloud, mobile—have not driven a comparable productivity surge, arguing these were largely deployment improvements rather than fundamental work redesigns. Turning to AI, they outline three possible future scenarios: (1) output explodes while jobs remain stable, (2) output stays modest but jobs collapse, raising productivity via a reduced denominator, and (3) modest output and modest job loss. They argue the second scenario is most likely, with AI displacing workers and boosting measured productivity without large output gains. The conversation then shifts to societal impacts: widening inequality between capital owners and displaced workers, potential labor‑market disruption, and the importance of timing. Finally, they offer practical advice for listeners and their own children—regularly learn and apply AI tools, stay on the “right side” of the AI divide, and track labor‑productivity metrics themselves. The episode closes with standard sponsor and disclaimer messages.
Chapters
Live from Schenectady, New York, it's the Metrics Brothers, Growth and CAC.
And I'm Growth, better known as Ray Reich, founder and CEO of Benchmarkit.
And I'm CAC, better known as Dave Kellogg, EIR at Balderton Capital, independent consultant, and the author of Kellogg.
And together we are the Metrics Brothers.
And we go together like inputs and outputs.
Like I/O to a disk drive?
Yeah, that's a form of I/O. Today I was thinking more about work, Ray. This is the Metrics Brothers after all.
Ah, like the inputs tell you what went into a product and outputs is what you got out?
Yeah, that's the idea.
Well, I like to live in the ratio between inputs and outputs. That's what really gets me excited.
That's right, Ray. And that ratio is productivity, right? How much you get out for what you put in. CAC ratio would be a— my name would be a productivity ratio, right? Here's how productive a dollar of sales and marketing spend is at producing ARR. So that ratio between inputs and outputs is what we want to talk about today.
And here we are. We got to what I wanted to talk about today, productivity, but not in a micro sense, more in a macro sense, since Everyone's talking about AI and the impact it's going to have on labor and the workforce. I thought that today we should talk about labor productivity.
Okay, Ray, this is a huge topic. I hope you're not biting off more than we can chew. But before we dive in, let's hear a word from our sponsor.
The Metrics Brothers is proudly sponsored by Benchmarkit, the B2B AI and software industry's largest, broadest, and most contextualized benchmarking index. Dave Kellogg and Ray Groth Wright have put the Metrics Brothers podcast on hiatus for remainder of 2026. All episodes we have ever recorded are still available on your favorite podcasting app. You can reach Dave at Kellogg and Ray on LinkedIn at Ray Reich. Thank you so much for listening to the Metrics Brothers podcast.
Well, Dave, we may be biting off more than we can chew, but, you know, labor productivity may be the most important metric in the US economy. And yet I think it's one of the least understood and one that can be dramatically impacted by AI. So I thought we'd cover 3 things today: how labor productivity is defined and measured, where those productivity measurements may fall short, And then getting into the meat, how technology has historically influenced productivity, and then the impact on productivity we can expect from AI.
And our goal today in 25 minutes is to take productivity from one of the least understood metrics to being one of the absolutely most misunderstood metrics for our audience. Just kidding. Our goal is not to do that. But here we go, Ray. Let's give it a try.
Okay, so the US Bureau of Labor Statistics, you'll hear it referred to as BLS, defines labor productivity as real economic output per hour worked in the nonfarm business sector. So basically, the equation for labor productivity is economic output divided by hours worked.
Okay.
Now, one of the most, I think, important things about labor productivity is what it excludes. So it excludes the output and the associated hours worked in farms. So they call it a non-farm measurement. It excludes government and includes household production, like household employees or owner-occupied housing services. So at the end of the day, only about 75% of our total GDP in the United States goes into this Economic output divided by hours, which is the US labor productivity.
Now, since I was on the board of an agtech company once, I wanna talk about why farms are excluded and why government is excluded. It makes sense to exclude farms because they're very volatile, literal seasonality, the original form, right, from the seasons. It's hard to measure. You have a lot of migrant labor, you have undocumented labor, so it's hard to measure labor hours. And today, it's only about 1% of GDP. So, it's kind of a difficult-to-measure volatile sector that doesn't contribute that much. So, they leave it out. That's why they talk about nonfarm. Government is big. It's a lot bigger than 1% of GDP, but there's no output measure. We can't look at number of cars sold, patients billed, dollars produced. How do you do that for a court ruling or an FBI investigation? You can't. So, government is excluded, as are the other sectors that you talked about. Yeah.
And the one other thing, even though it could be assumed with our listening audience, right, it does include white-collar employees. And even though they use things like employer payroll records and some household employment surveys to get the number of hours, it's not hardcoded to 40 hours for white-collar employees, but it is in a range of 38 to 42 typically. And the BLS will say, but it excludes overtime for exempt employees. So, hey, Dave, even if you're working 70 hours a week, Right? They're not going to count that if a lot of it was at night or on weekends. It doesn't differentiate an hour of admin work versus high-value output work. And it does not include the impact of non-human labor. So it doesn't try to factor in, well, robots worked 1,000 hours in that plant. It's not included. Yeah.
So for so-called non-exempt employees, I think this is a much easier metric to track because that's how you're paying people, right? They're not exempt from overtime regulations. So you track Their hours. For white-collar employees, how elitist of you, Ray, nobody tracks their hours. So I don't know how BLS could track. You could ask a company how many hours their white-collar employees are working and they wouldn't know. So how in the world would BLS know? So I think it does not count, notably, we can call it overtime for exempt employees, right? Like nights, weekends, all the stuff you said. It doesn't include the value of the work or the quality. So an hour worked by the The receptionist is the same hour, same as an hour worked by the CEO. Hopefully, the CEO gets more done, but who knows? And it doesn't include, I know this is your favorite, nor would I expect it to include, but let's be explicit, it doesn't include non-human labor. So if we replace somebody who is working 40 hours a week with an agent, there's no notion of substitution involved. That just simply those 40 hours go away because they're not attempting to count non-human labor. Go ahead.
And that's going to be one of the interesting impacts on labor productivity of AI, which we're going We'll talk later. Now, just real quick, just from a historical kind of trend line, you know, if you looked at 1980 before the big, you know, IT boom, you know, economic output per hour worked was around $60. That's US dollars. And today it's around $104. So from $60 to $104. So what is that, like almost 70% improvement over the last 45 years, Dave?
Yeah.
Yeah, yeah. I'm just gonna round the trajectory and say roughly we've doubled it. We're on track to double it in '50, right? So there have been material increases in productivity in my career. I started working roughly in 1980, a little bit after. But so yeah, so productivity has gone up. It's kind of slow and steadily. We'll talk about the growth rate in a minute. But before we do that, Ray, maybe just very quickly, let's talk about how our productivity compares to other people's. I'll take the lead on that. We have some OECD data. D data here now. It puts the US at $112 per hour. So roughly in alignment with the data you showed. But the number 1 country is, ta-da, Ireland, $140. Number 2, Luxembourg. Number 3, Norway. UK, $87. France, $91. These are all kind of tier 1 countries. South Korea, $80. Tier 2 runs here from memory, like $35 to $45 in the kind of middle tier. If you looked at the 50th ranked company, It would be in that range. And if you looked at the bottom decile, it'd be more like $2 to $8 per hour.
Yeah.
And by the way, just real quick, this has to be normalized on a global basis. And that's one of the things OEC does. So they use something called PPP, which is purchasing price parity. A little bit of a rat hole, but that's how we get the global comparisons. It's not something that our US Bureau of Labor and Statistics does.
Yeah, I think it's purchasing power parity is my guess. But yeah, they're trying to say, based on the cost of a basket of goods, how far does your money go in country 1 versus country 2 and kind of normalize for that. So it gets complex when you compare across country, but I think it's important to understand that we are a leader on this metric, but we're not number 1, right? We're quite a bit behind Ireland. Now, they have some advantages we could discuss, but we shouldn't. But we're— let's just say we're a world leader in productivity at around $110 per hour. Wow.
We are. So let's move to something that we often hear about if you read the Wall Street Journal, listen to CNBC, and that's productivity growth. And this just really measures the period-over-period increase in labor productivity. So it's a percent change in output subtracting the percent change in hours worked. Simple example, output grows by 3%, hours worked grow by 1%, you subtract those, productivity is 2%. And what's interesting here, Dave, is You know, this really kind of peaked in the 1950s and '60s, that post-war boom. It was almost, you know, up to 3.8% to 4%. A little choppy in the '70s and '80s. Then we saw a strong bump in the mid-'90s to 2005, kind of that IT revolution where information technology transformed so many jobs. But it's been pretty soft as far as productivity growth over the last 20 years, kind of in that 2.1% range.
Yeah. I mean, if you look maybe a little more broadly, I see 2 big buckets there. I see a bucket from 1950. We're looking at some charts we had here, but a bucket from 1950 to 2000, where productivity was growing at around 1.8% and GDP was growing at 3.5%, right? And then I look in the more recent history, 2000 to 2020, I see productivity down, growing at 1.2%. So, a 66% relative decrease in productivity. So, it's down.
Yeah.
And also, because the labor force also— labor force growth also went down quite a bit from 1.7% to 0.4%. It means GDP growth is growing in the 1.6% to 2.0% range, which is below our historical average of around 3%. Long story short, Ray, I think the story here is from 1950 to 2000, thanks to kind of factory automation in the '50s and '60s and the IT boom of the '80s and '90s, We saw real productivity growth of 1.8% and strong GDP growth of 3.5% above the 100-year average. And in the last 20 years, eh, not so much.
Okay.
So Dave, you know how many of the people in our audience have worked in the tech industry for the last 20 years? Hey, we helped create SaaS and mobile and cloud. And the question is, after trillions of dollars investment in all those technologies, over the last 20 years, why is productivity going down and why is GDP growth lower? So I've got some ideas here. So number one—
Yeah, and just to be clear, Ray, productivity is not going down. Productivity growth is going down. So the rate of growth in productivity has decreased from 1.8% per year to 1.2%. Keep going.
Yeah, so that economic output growth was, over the last 20 years, was lower than the average over the last 70 to 80 years. Labor participation rate, that is, for the total amount of people who are in working age are actually being employed, it was 67% 20 years ago. Today, it's only 62%. So that means that it's a 5% lower participation rate in meaningful employment. And with productivity slowing due to the SAS errors you talked about, and for me, I think it's because in the 1980s and '90s, Digitization actually transformed work. But over the last 20 years, we've just improved some of the efficiency of digitization, but we're not fundamentally transforming what's being done. And for me, I think that's where AI is going to be different. It's going to once again fundamentally transform what people are doing, not just increasing efficiencies at the margins.
Yeah, so 2, 2 things to unpack there, Ray. One, Yes, I think the consensus from our research, and I was unaware of this, by the way, that the average GDP growth is around 3% over the last 100 years. And we've been only growing at 2.1%, i.e., sub-average GDP growth, despite kind of the SaaS boom and the cloud boom that we've lived for. So, it was something that was a surprise to me, especially with the stock market, which has been booming, right? And they always say the economy is not the stock market. Well, it's not.
Yeah.
So, that was the first thing. I was kind of surprised to say, wow, we're living in an era of substandard productivity growth. And I do buy the argument that you made that said, look, in the '80s and '90s, moving to computers basically fundamentally transformed the work we did, right? We went from having no CRM to having a CRM that changed how the work was done materially and how we tracked it and analyzed it. Moving it to the cloud, and I've always felt this way, it was just a better deployment model.
Yeah.
I was like, I often felt like, literally, I had a roadmap in 1998 for my product, and it got put on hold for 5 years to go rebuild the product for the cloud, right? And that was, as a product manager at the time, that was the way I felt. I'm like, shit, this cloud thing is bad because it's setting me back functionally, but by a very long time, because I have to blow up— remember, Business Objects was originally a PC tool, so we literally had to blow up every line of code and create a new server version of Business Objects because the cloud was here. So, the argument that Ray was making is that, you know, in the '80s and '90s, we basically gave people systems and automation that they never had before. It was all on paper. So, that drove a big productivity boost. And the cloud was just more incremental. For us in the industry, it wasn't incremental. This is like a radically better way to deliver software. But the software, as we both know, was kind of radically the same, right? Like Salesforce was an anemic version of Siebel, right, for a really long time.
I'll give you another example that I live in every day. I started using Excel in the early '90s, and Google Sheets, even though it's delivered via the cloud, doesn't provide me any additional value over the tool I've been using for 30 years.
Right. So, this touches on, and we're gonna get slippery sloped here in a second, but what I think it meant to me was because we weren't really redesigning the work, we just had a better deployment model, a more cost-effective way to deploy software, The software industry did well, it grew. And I think companies, I'm not sure about this, were able to get more profitable as a result of that growth. But that profit doesn't turn into labor productivity. It turns into stock market going up. Hey, companies are more profitable, that's great. But that's not correlated to labor productivity. I don't wanna say unrelated, but we were able to be more productive because we were spending less on infrastructure that did roughly the same thing. So first, it's an incremental improvement.
Yeah.
And second, the result was more profitability, not more productivity in the sense of more output, right? Growth was actually relatively slow, at least in the software sector during that timeframe or parts of that timeframe. Okay, so here—
Go ahead.
Here we are, right? For the last 20 years, the SaaS mobile cloud era didn't dramatically improve labor productivity era. But now we're all excited about AI and it's gonna change everything, Dave. So I thought maybe we could talk about that a little bit.
Before we do that, Ray, let me do one analogy I like, which I think is relevant, which is a lot of problems we face, including this one, are what I call same-store sales problems. And enterprise software is a same-store sales problem. Let's be clear. What is same-store sales? Pretend you're Starbucks and you're reporting your quarterly earnings. You report 2 numbers. One, how much overall sales grew, and then 2, how much same-store sales grew. And same-store sales is for the stores that were open 1 year ago, how much did they grow? I.e., let's decompose our growth into 2 things. How much more productive did the existing stores get, same-store sales, plus effectively how much revenue do we get from the new stores that opened, right? And that metaphor, we don't really have a same-store sales metric in enterprise software. I think we should. That's the same thing as GDP growth, labor, and productivity, right? You're basically saying, If GDP growth is going to be 3.5%, I'm going to get that by kind of combining productivity, i.e., same-store sales up by 1.8%, plus labor force, i.e., number of stores up by 1.7%. And the only reason I mention this is it's how we think about GDP. It's how we should think about enterprise sales team. And I just think it's a really useful metaphor. And if you want to learn more about same-store sales, it's a well-known metric in retail. You can Google it.
I love that metric. And by the way, it's funny, I always compared net revenue retention to same-store sales. It's how can I sell more from the same kind of location, same products per se.
Think about that. I'm going to actually argue it's same-customer sales just to split hairs and be disagreeable. I don't know what store they bought at, but, but, but yes, NRR is kind of same-customer sales. But, uh, in any case, on we go. Because same-store sales, a store in enterprise software, Rep is a store, right? Let's just be clear. What's the store in an enterprise software company? Joe is a store. Al is a store. Mary is a store, right? It's same rep sales in effect. Okay. Back to your question, Ray, which I think is the big question for the episode, which is, okay, now that we've had this primer on productivity, what's AI gonna do to it?
Yeah.
Well, I'm going to couch this in kind of 3 high-level potential scenarios, and then I'm gonna go into the one I think. So we could see Output exploding. All these AI agents are finding new ways to get new revenue, etc. And jobs are fairly stable. So productivity would be up a lot because output explodes. That's not the one I think is going to happen. Output could be stable, or i.e., GDP, economic output could grow in that 2-3% range, but the number of jobs collapse. The input hours collapse. So productivity would be up a lot, but it's a much uglier transition because a lot of people are out of work. By the way, I think that's most probable. The third option is output could go up modestly, kind of that 1.5%, 2%, 2.5%, and jobs are just down modestly. I don't think that's going to be what we see over the next 3 to 5 years. All 3 of those scenarios raise productivity, but the one that I'm If I had to bet, I would say I think the number of jobs being displaced is going to increase productivity, not significant output increase.
Okay. So Ray's arguing that in fact the denominator is going to go down, not the numerator going up. Or to the extent we see an increase in productivity, it'll be more denominator-driven than numerator-driven, right? That's what Ray's saying. My take is similar. I mean, first, I do think there will be strong GDP growth ahead. I think productivity, like typical productivity growth, just productivity growth, again, from 1950 to 2000, it was 1.8%. GDP grew 3.5% per year in that time period, but productivity contributed half of it at 1.8%. The other half came from the labor force growing at 1.7%. I think in the coming future, I think we could get productivity growth of 2% to 3% just from productivity. And the labor force is growing much slower now. See all the immigration policy things, like labor, both participation in the labor force and unemployment and immigration policy. I think we may live to regret some of that because one thing that grows GDP is more people. And when we don't have more people around, it's an issue. So, I think productivity is going to be boom. I'm a big believer that AI is structurally remaking work. It is not just a better deployment model like cloud was, let's just say. It's going to help fundamentally change work. Some of that will be unrecorded, right? It just means that if white-collar workers work 50 hours instead of 60 hours, it won't show up in this metric, right? But some of it will get recorded. And I do think there's going to be I don't want to be dramatic, but massive, heavy labor market disruption as we get from here to the other side of this, right? CEOs talk about cutting 30% of their company, 50% of their company. You hear very large numbers when CEOs talk about what they think they'll be able to cut from AI. I think 2 things. One, those numbers are optimistic from the CEO's perspective, happily for society.
And second, Oh, shit.
I think the numbers are too high. Oh, what was the other point? Wait, I cannot remember the other point I was going to make.
They're too high.
Oh, and the question, I guess, is going to be about timing. Like, if they hit all at once, it's going to be highly disruptive. Whereas if they happen over time, it'll be less disruptive. But I personally do believe we're in store for kind of a lot of societal disruption and societal unrest as we go through this transition because companies are going to try and eliminate a lot of jobs. I think productivity is going to go up, but no one's going to care because if you have a job, you're happy. And if you don't, you're going to be unhappy. You don't care how productive the other people are, right? So, that's my take here, Ray.
Yeah.
And something I know we talked about as we were preparing for today, this could really further create a bigger chasm between the haves and have-nots. Because if companies are earning more, investors are going to be happier. But if employees are getting displaced, or making less per hour, they're not going to be able to spend as much. So it's a very delicate balancing act between the people who are capital kind of investors and the people who are hourly employees.
Yeah, I think so. And let's be clear, if productivity makes your productivity go up, you can get paid more. However, if your job gets eliminated and you have to get a lower-paying job, then, then your pay goes down, obviously. But I think we'll see both effects, right? If you get bounced out of a job, then you'll take whatever job is available. However, I think you're arguing for kind of this is going to create a divide, and not as if we were in need of one, but I think it will further the existing divide. And then I think the time ahead could be pretty disruptive, but I think GDP is going to grow well. I think productivity is going to grow well. I think the statistic we're not talking about is unemployment. And I think unfortunately, that's going to grow well. And it's gonna create a lot of problems. And I do think it will drop. I don't think we're headed to permanent unemployment, but these transitions are really hard. We should convert this to advice for the audience before we wrap up, Ray. Given everything we seem to believe, like we both have kids who are 21 years old, how— what do we tell them?
Well, after I tell 'em, don't depend on your inheritance. I actually have encouraged all 3 of my children. I said, once a month, Learn a new AI tool. Let's apply it to something you're working on right now. Let's do some research leveraging an AI tool. So I do believe that the people who know how to leverage AI tools into real-life situations are going to be highly employable over the next 5 to 10 years.
Yeah, I think I had similar advice, which is you want to be on the right side of this. If you see a divide coming, there's a wrong side and a right side. And the right side is, you know, a lot about how to use AI practically to get get a job done more efficiently, that's a winning proposition. So, that's my first piece of advice is make sure you're on the right side of the divide. That applies to all of us. You know, both Ray and I use AI tools extensively because we want to be on the right side. Jason Lemkin has like rewritten half his company in vibe coding, right? And cut his sales force from 8 people to 1, right? So, he's practicing it. You want to be on the right side of the divide. And the other one to me, to be honest, is a Grateful Dead song called Keep Your Day Job. And it's like, this is not a great time to quit your job in a huff, in my opinion. 'Cause you're mad at your boss. I mean, you know the labor market for your position, but certainly for my kids, especially the younger ones, it's like, this is a time to, if you're in a good place, stick with it. 'Cause finding jobs can be hard. Finding jobs, especially for entry-level people right now.
Totally agree. And I would encourage everyone, just start tracking labor productivity. You can see if Dave and I are kind of directionally correct here. An example, Q3 '25, Labor productivity was up 4.9%, about 5.4% increase in output and about a negative 5% decrease in hours input. So it's going to be interesting these next few quarters, Dave.
It will. Okay, well, that— let's wrap it up, Ray. Thanks for the episode, and we'll see everybody next week.
Goodbye, everyone.
The Metrics Brothers AI, SaaS, and More is a production of The Metrics Brothers Growth and CAC and a member of the Benchmarket Podcast Network. By accessing this podcast, you acknowledge that the Metrics Brothers make no warranty, guarantee, or representation as to the accuracy or sufficiency of the information provided or the humor content of the jokes told.
Ahem.
Ray? Information, opinions, and recommendations presented are, according to our spouses, probably wrong and provided for information purposes only. This podcast should not be considered professional or, for that matter, unprofessional advice. We disclaim any and all liability for any direct, indirect, indirect, misdirect, incidental, special, ordinary, consequential, inconsequential, or other damages arising out of any use of, or God help you, reliance upon the information presented here. Ray Grothreich is based in New York City and available on X at Ray Reich. Dave Kellogg is based in Silicon Valley and available on X at Kellogg. Schedectity, which is French for unspellable, is not our actual production location.
Thanks for listening.
