Will AI solve the productivity puzzle? With Nick Bloom

0:00The story of the 2010s was one of disappointing productivity growth. Today, that story seems to be changing. In the US, productivity growth seems to have perked up a bit, while there is deafening yelling about how artificial intelligence is going to turbocharge the economy. Inflection points in the path of the world's growth frontier are rare, but according to my guests today, we could be at one.

0:33This is The Economics Show with Samaya Keynes. I'm joined today by Nick Bloom, Professor of Economics at Stanford University. Now, Nick isn't officially yet a friend of the podcast, but he is certainly a friend of my column, as I have written several drawing on his research. And I'm just delighted that he is here to join me today. Nick, thanks so much for being here. Samaya, thanks for having me on. I've definitely listened to a lot, but yes, So I've been on the other side of many of these podcasts. Okay, well, you've listened, so you know the drill. On a scale of 1 to 10, where were you a couple of years ago when it came to productivity boosterism?

1:12How optimistic were you feeling? So 1 is down in the dumps. We're going to be in this kind of low productivity era forever. But 10 is, okay, well, you know, chat GBTs come out. I'm excited. I think this is going to deliver the singularity. A couple of years ago, probably a two or three. And I don't know if you're going to follow up and ask, but I'll go there anyway today, maybe like a six or a seven. So I've shifted my mind a lot in the last 24 months. Okay, interesting. Well, yes, I do want to get to that. But let's just start with your pessimism a couple of years ago. So why? Well, it's kind of an empirical thing.

1:52If you look at the US, and in fact, there's data on most of the OECD, what you see is productivity growth has been slowing. It's been on a kind of downward trend since the 50s or 60s. So if you go to the US, for example, in the 50s, you saw productivity growth of 3%. It's slowing to kind of 2 % by the 80s, and it was falling to closer to 1 % just before the pandemic. And at the same time, we're spending ever more money in R &D. There's more scientists, there's more activity in universities. And so those two things look kind of depressing. Working harder, doing more research and productivity growth is slowing down.

2:26So that was my pessimism as of a couple of years ago. Okay, but this is interesting because you wrote about that in a paper published in 2020 called, Are Ideas Getting Harder to Find? And your answer then seemed to be yes, right? It looked like we were running harder to stand still. or your analogy was, you know, if you think of ideas as apples on a tree, maybe we'd already picked the lowest hanging ones and now we were having to spend more and more energy reaching the highest ones. There's actually a recent paper that's kind of challenged some of the conclusions in that study, suggesting that actually, rather than ideas being harder to find, it could have been that ideas were simply getting harder to share, right?

3:10So there was something about the diffusion of ideas that were slowing down. Have you read that research and what do you make of it? Yeah, so that's a great paper. So just to be clear, I had a paper with Chad Jones, Mike Webb and John Van Riena and it's kind of like an input output thing. Think of the inputs into innovation. It's all the money we spend on science, scientists, labs, etc. And the output is productivity growth. And it looks like that relationship is getting worse and worse. We're spending ever more and we're getting less productivity growth. And the question is why? And we thought, well, maybe it's ideas getting harder to find.

3:46It's like the coal mine or gold mine type thing. You're having to dig further and further into worse and worse seams. I don't think it's so clear to us why it was slowing down. It was just very clear that growth was just becoming ever tougher. I should also say that isn't what we see going back over the entirety of history. So if you go back to about 1760, 1770, the Industrial Revolution starts. And before the Industrial Revolution, just really drawing on British data, productivity growth is close to zero. I mean, in the 1700s, they weren't a lot richer than they were in 1000 AD. And that's a huge period of time and very little growth.

4:21And then the Industrial Revolution starts and productivity growth actually starts to pick up. So for a couple of hundred years, it looks like ideas were becoming easier to find or easier to spread. And it looks roughly the inflection point was the 1950s. That was kind of peak. And then it seemed like it had been on a downward slope from then until at least until recently. Okay. But I suppose the model that you were building in that paper was that this was just something fundamental to the nature of science, right? We found lots of ideas and maybe we'd run out now. And therefore, we weren't going to grow as quickly as some had hoped.

4:56There are two possible challenges to that. One is, okay, actually, maybe you need to look more closely at the mechanism, right? Is it diffusion of ideas or is it just finding the ideas in the first place? And that's where this new evidence, I think, sheds some light. But then there's also the bigger challenge of AI, right? Because clearly here we have this idea that's pretty big, where it does seem that there's a lot of scope for productivity gains. So, you know, how do you think of AI within that framework? I think AI could fundamentally change what's happening. So in a different paper with a group of folks across many countries, actually, we've been surveying thousands of businesses, in particular focusing on CFOs and some CEOs who kind of know what's going on and asking them, what's the impact on AI on employment and productivity of the last three years?

5:46And what do you predict the impact will be on the next three years. You see execs in the UK, US, Germany, Mexico, Canada, many countries report that looking back, there hasn't been much of an impact on AI on either employment or productivity. And basically it's there, but it's just not widely diffused yet. But looking forward, so looking forward to 2029, they are pretty optimistic. And so the average number they're reporting is it's going to boost productivity growth by about 1 % a year. And that is, you know, It's between three quarters to one percent a year. That is an enormous number. So it's nowhere near Dario Amodi type numbers.

6:22He's talking about 10, 20 percent and, you know, flying cars by the end of the week. But on the other hand, it's a huge material change. And my view is kind of I'm about 50, 50 at this point now that productivity growth is genuinely changed direction is going to start heading back up again. OK, so you're doing these surveys of CFOs and they're telling you that great things are coming. could this just be part of the ai bubble everyone's optimistic so these are panels of cfos we've been serving now for more than 10 years their numbers and employment on sales are incredibly accurate so much so that they map aggregate gdp employment numbers and it's used by the fed and the mpc in the uk for example in monetary policy we also actually asked them to forecast sales employment investment and you know you can look a year later at their accounts and they're pretty good.

7:14So on everything else, their track record is close to impeccable. Of course, AI is a new tech, so it's hard to call. But basically, we asked them looking forward three years because they know what's going on in the short run because they see their businesses. If I wanted to know what AI is going to do 20 years out, I'll go speak to a technologist. If I wanted to know what AI is going to do three years out, I probably want to talk to someone that runs a business because it's really in the short run all about bottlenecks and implementation. So on that, I think, you know, this is the wisdom of crowds and it's probably where I put my money.

7:45How diffuse are these gains, right? Or is it just a handful of CFOs and superstar adopters who are expecting loads and loads of productivity growth? How broad could these productivity benefits be? About half of the companies report positive numbers and about half report zero. So another metric that was really interesting as we asked in the US, two, three months ago, how much money are you spending on AI per person? And the average was a couple of thousand dollars, which is a lot. But the median was zero or close to zero. Most firms are not spending a huge amount, but there's a few that are spending enormous amounts on AI.

8:24But it's concentrated in certain sectors, so tech, finance in particular, business services. And then if you go to something like construction, I mean, I had my house extended recently. It was a very expensive project hit by a lot of regulations and AI was nowhere in sight. So there's certain industries I think it's really, as of yet, having very little impact. So you've done some research looking at how technology diffuses around the economy. Are there any lessons from that research that we could apply to today? Yes. So in another paper with Akash Kalyani and co-authors, we look at text data, and particularly in job postings, earnings calls, etc., to see how new technologies going back 40, 50 years spread.

9:07Interestingly, what you see is new technologies initially turn out to be very geographically concentrated. So think of stuff around the internet. It really formed in Silicon Valley, this stuff spreads out, but it takes 10 to 20 years, particularly for low paid jobs. So the classic is new technology comes in, ton of employment, but in kind of college grad, high skilled jobs and locally. Over time, it spreads out. So technologies definitely drive growth, but the initial impact tends to be concentrated. And the reason I think this is important is it's politically sensitive. So politicians want to drive new employment, want to drive growth, but it's harder to defend it if it's basically high paid people in certain parts of the country.

9:47Right. And that's clearly a dynamic that could play out today with all of those very well remunerated tech bros. Okay. A lot of what we've been discussing so far is really focused on the US. I mean, you're talking about the CFO survey, which is very international, but are you seeing any differences across countries outside the US in particular? Yes. If you look at AI adoption, the US is highest. Canada, UK, Australia is not far behind. Germany was lower. We have data for South Korea is actually pretty high. We have data for Mexico. South Africa is a bit lower. It looks a lot like the GDP gradient.

10:22I mean, on AI in particular, it is, for modern technology, it is incredibly skewed, and particularly the value creation into a few parts of the country. I mean, as you say, I think this is a major obstacle for AI, is that most of the value right now is going to investors and a few folks in Silicon Valley, not far from where I live. And of course, there's a backlash. And the backlash is partly over data centers and energy prices, and also partly more recently over, you know, AI escape and terrors of things like Terminator-type movies. So for AI, the industry, I mean, one of the most important things is to spread the wealth and spread the employment as fast as possible, because otherwise, you know, I can see it just being regulated out of existence or at least, you know, really slowing uptake.

11:03Does anyone outside the US have the capacity to do that, though? I don't know. AI has this issue, and as an economist, it looks a lot like a natural monopoly. So there's incredible returns to scale. And when there's incredible returns to scale in an industry, you typically get two or three big players. And for these fundamental models, you're seeing in just a few big players, Gemini, GPT, Claude, et cetera. And that is not a healthy dynamic because you get a few companies dominating the industry. Right now, they mostly, but not entirely sit in the US, you know, some in China. Coming back to our work on technologies, it's harder to generate broad support for technology of all the benefits and all the wealth accrue to wealthy, you know, high educated folks in small areas.

11:47And it just doesn't spread out. You were recently at the National Bureau of Economic Research's Summer Institute, organizing a session on macroeconomics and productivity. This is sort of like summer camp for economists. I'm assuming there was a lot of chat about AI. What's the mood in the profession? How is AI affecting economists? AI reminds me of this quote by Bob Lucas. So Bob Lucas was a Nobel Prize winner, has now recently passed away. He had this saying, which is, when you start thinking about growth, it's hard to think about anything else. And now I've replaced the word growth with AI. It just feels like the profession is becoming dominated with interest on AI.

12:33I mean, another trend I'm noticing is people joining AI companies. I have some amazing students that have joined it. Many brilliant colleagues are joining some of the leading AI companies. That's a notable trend that's really accelerated in the last year or so. So yes, over lunch, over dinner, in sessions, I would say 25 % of the discussion is probably about AI, and that seems to be rising rapidly. I think economists are kind of in the middle to slightly optimistic. I mean, there's a big spread, but, you know, Modi-type figures are putting out incredible growth rates, some of the Silicon Valley bros.

13:06I mean, the odd thing about them, by the way, is they're also talking about incredibly negative effects in employment. Economists tend to think that, on average, there are likely to be some maybe negative effects in employment, not massive, and some positive effects on productivity, but we're not seeing them there. I'd say they're a moderating force on some of these views. Maybe they're not trying to sell equity in their startup. They're trying to get it right. So that's probably a positive on their side. What about the actual productivity of economists themselves? Is it massively increasing the amount of research or even the quality of research that people are producing?

13:40The quality of research and the speed has just exploded. So literally this morning, I have a paper very appropriately on AI with Darren Asimoglu and John Van Rienem. and the three of us were meeting and we're working with a group of students and pre-docs on it. And it was just incredible. I literally made the comment, you know, you guys have done three months of work in like one week. It was astounding. Can I get a sneak preview of what this paper with Daron Asamogler and John Van Wienen is about? Yeah, absolutely. So what we're doing is we're trying to think of the causal effect of AI and firms productivity.

14:17And we have looked at what directors of companies are saying on LinkedIn. and you can look at whether individuals are like pro, neutral or negative in AI. I mean, look in particular at those directors moving across companies. What we see so far is some positive effects on employment. We haven't managed to pull together the productivity effect, but that's kind of the thought process. And there are a few other studies. There's one can ramp recently whereby repeatedly in this micro data, you tend to see firms that adopt AI see increases in employment. The problem is obviously over causality. It's like, look, there's two stories.

14:51One is AI is great for business and you expand and you go and hire a lot of employees and that's very positive. The other is I'm a firm, some huge customer order comes in or something good is on the horizon. I decide now's the time to invest in AI. And it's really hard to pick those two apart. But at least on trends data, what you can say is firms that are investing in AI do seem to have increasing employment rates afterwards compared to firms that don't. Okay, one more question on economists. have you seen any economics slop uh so i'm talking about you know researchers just using ai to write whole papers that are essentially rubbish i'm not sure i'm gonna name it there are certain individuals that seem to dramatically accelerated their output of papers whether they're rubbish i don't know i haven't read them in detail the abstracts look fine but i haven't read the full paper.

15:44You know, it feels like for me personally, working with a bunch of younger folks, it's much easier to run analysis. I think what's going to happen is actually the bar is going to be harder to publish. But yeah, I do worry, you know, the thing that I worry about is some paper being published and two years later discovering some horrendous bug in the code and it's because AI has written it. And yeah, I've seen that when I've been looking through stuff. Sometimes you've asked questions and, you know, you sit through with your colleague and they'll go through the code and there's something odd and it's because AI is misunderstood.

16:11It's not that AI's done and made a mistake. It's just, you know, humans are vague and they're not precise and it interprets it the wrong way. What about teaching? Has it affected the way that you're interacting with students? It's just had a massive effect. So basically, any written material that's handed in or even presentations that people give in class is going to be generated by AI. So for economics, you know, even including graduate teaching, it's made it harder slash different. So I focus much more on class discussions, like take home coursework. You know, it's just not something you can grade anymore.

16:46Friend of mine was saying, just saying in her maths course, she'd never seen such high scores for a weekly class test that students take home. And then in the exam, they bombed it. And obviously what was going on is the students were getting AI to fill it in. So there's much more focus now on in-class activities and things that AI can't do is generally a big challenge for universities. And I would say one third to one half of the discussions around teaching is how to deal with AI. I mean, it sounds like a sort of reduction in productivity of teaching because it's so much harder to assess now. Are there any other areas that you've heard of where actually AI could also be reducing productivity?

17:25You know, another example of AI having negative productivity effects when I talked to heads of HR is on interview. So a lot of tech firms used to hire folks based on coding tests. They can't do that anymore. Now I've talked to a few and they say even in interviews, they will be online. And they're figuring out folks are using AI prompts and they're just answering the questions. So they're having to call people physically into the offices. Quite a few big famous tech firms are now saying they're running interviews in person in final rounds and even middle rounds because AI, otherwise they don't know who you're interviewing.

17:56Are you interviewing Claude or the person? okay well we should throw to a break now but when we get back i'm going to ask you about another topic that you have done a huge amount of research on work from home

18:18for years credit scores were generated with just a snapshot a snapshot that left millions of people out of the picture. So VantageScore is modernizing the credit score. VantageScore uses 400 % more credit report and alternative data than legacy models, bringing more people into the credit economy, like 5 million more mortgage-eligible people. Banks win and francs win. Lenders win, and so do Lindas. When you build credit scores the right way, everybody wins. Get started at VantageScore.com. We are back from the break. I want to ask you now about a LinkedIn post that you did quite recently. So we were talking earlier about what has happened with productivity growth recently.

19:05You're skeptical that it's large language models. Is there something else that it could be? I posed, could it be working from home? So initially, you'd be hugely skeptical and think, how on earth could that be? Well, A, there's a bunch of studies now that if you let people work from home one or two days a week, you can see big increases in productivity. Where does that come from? The reason you can get this is, one, it's actually quieter at home. So think about you're trying to study, concentrate, focus. It's often a lot easier to do that at home. And some share of every job requires that. And B, it saves a lot of time on commuting.

19:42So the typical office worker is saving about an hour and a half every day they work from home on commuting and actually less grooming. Turns out people spend less time grooming, including quite horrifically, less like to brush their teeth and other things I should probably not talk more about. Anyway, that means you can potentially have productivity effects from work from home. You can also have labor supply effects. So I have a paper showing that in the US, there's about half a million people with a disability that look like they're working now that wouldn't otherwise be working because of work from home.

20:14And you combine those together, you can potentially see positive productivity effects. That may be one of the things driving this boost from 22 until now. What was the reaction to that post? Did people agree, disagree? Well, the reaction, I think, was generally very positive. But this is very much who's following me on LinkedIn and people who are following me are people who are interested in work from home and they tend to enjoy it and like it. I would say 80 percent agree, 20 percent disagree. Some of the strongest disagreement came from folks selling corporate real estate who don't like it. I'd say at this point, my sense, by the way, is the much bigger effect of work from home is actually labor supply.

20:54So if you're America or the UK and you want to boost growth, there's two ways to do it. One is to make each person more productive, and the other is to have more people working. And work from home, I think, has a much bigger effect on the second, because what we see in the data is a bunch of folks that live far from the office, that have some disability, have care responsibilities, maybe a studying or close to retirement, are working longer and more hours and are working just full stop because it's easier if you have to commute in every day. and that probably is the bigger effect and that could add 1, 2, 3 % to GDP.

21:29But would that add to productivity? Because productivity is output per unit of labor, right? So if you're just bringing more people into work, if they're below average productivity for whatever reason, then that could actually reduce productivity, no? Exactly right. So how that adds to, I should carefully state, measured productivity is not clear. So how do we get measured productivity? you look at value added per hour worked. And yes, if you just brought in a bunch of extra people and you perfectly measured their hours of work, you know, it depends on whether they're above or below average productivity.

22:01You may think of their marginal employees, they're probably less productive, so drag it down. The tricky thing is measuring a lot of this is fiendishly hard. For example, if you look in the data, many people report working 40 hours, but clearly most of us don't. You know, we're probably working many more hours. And if people save an hour and a half a day on commuting and spend some of it to work more, that typically isn't measured and it just shows up as higher output. So breaking down GDP growth into the bit there's productivity and the bit there's hours is tough. My personal view from what I've seen in the research is work from home probably boosts the hours component a lot more, but it may show up in the data as productivity.

22:37Okay. So there are these two effects of work from home on your output, right? You don't commute, but also you might be able to focus. It feels like there are many employers for whom this is still a challenge tricky subject, right? There are still many employers who want their employees to be in the office more than their employees want to be in the office. And that's sort of a puzzle if working from home does deliver productivity benefits. What's your explanation for why there is that mismatch? Or have I misunderstood it? No, totally right. So two things. One is a puzzle, one is less of a puzzle.

23:16So the less of the puzzle is employees like to work from home more than their employers want. You know, employees generally, if you look at the data, they want to work from home roughly half the time. So why would anyone want to come in at all? When you interview people, they're like, I prefer going in because it's good for mentoring, it's good for socializing, you know, I feel isolated at home. So it's not like everyone wants to be fully at home, but typically it's about half-half. When you look at employers, they typically want people in three and a half, four days a week. So there is a gap.

23:45And in some ways, it's natural. Employers probably want their, you know, workers to work more hours and workers want to work less hours. I think what is harder to understand is why there are employers forcing folks in five days a week when it looks like that is probably not a good decision. So take, you know, the mythical Goliath National Bank. If you've ever seen the TV show How I Met Your Mother, that was kind of the make-believe bank that Barney worked for. So, you know, Goliath National Bank may force its employees in five days a week. Why would it do that? I struggle a lot to understand. You know, once people are in four days a week, that's a lot of face-to-face and mentoring and connecting up, but he may think the fifth day, you can let employees say on a Friday, work from home, catch up on quiet, reading work, etc.

24:29There are various stories as to why CEOs have people come in all five days. One is, you know, they feel uncomfortable. They can't see what folks are doing when they're working from home. They struggle to manage them. Another is, you know, there's a kind of power struggle going on. There was an entertaining article in the New York Times recently claiming that it was about CEO egos being massaged to force people back in. A third story, CEOs may be right. Look, our research may be wrong. I should say, if you look in the data, what you see is about three quarters of Fortune 500 companies let people work hybrid.

25:01The other quarter are mostly fully in-person and about 7 % are fully remote. So there is a mix. I would say, you know, there is definitely no one size fits all in corporate US, UK. Yeah, I mean, I guess it's possible that even if you're finding productivity effects at the individual level, you know once you normalize working from home say maybe two days a week there are some people who just take the mickey and really not come in very much at all and then you know individually you might still be able to find this this productivity effect but what you're doing is for the organization as a whole could be weakening say the culture or you know making it hard to make fast decisions so there may be this kind of aggregate productivity that is held back do we have any evidence on that kind of phenomenon?

25:47Certainly there's evidence that you want to have some days in. I just put out a paper called the value of one day a month in the office. So we do a randomized control trial in a company whereby people are randomized into either being fully remote, it was a company that started off being fully remote, or coming in one day a month. And what you find is employees that come in one day a month are 8 % more productive, and that grows slowly over time. And also their quit rates fall by a third. When you talk to them, interview them, they say, look, when I come in, it's easier to learn, look around. I don't need to come in every day.

26:19But when I come in, I talk to my colleagues and it feels more connected. So my sense is you don't probably want to be fully remote for the typical job, and you probably also don't want to be in every day. Best practice looks like, it varies a lot, but maybe having folks come in, say, three days a week, and importantly, making sure they come in on the same day so they connect up. It's also true that better manage organizations and more able to handle work from home because they have a good sense of what employees are doing. So think of a very disorganized company. You kind of need to physically see people to know what's going on.

26:51Think of, you know, kick-ass, well-managed company that has great performance reviews, 360s, etc. You don't need somebody in every day to physically see what they're up to. Okay, one last question. You've done a lot of research looking at working from home. Are there any unexpected effects of this huge societal shift that we've experienced since the pandemic? I think in some ways the most interesting, I'm not sure it's unexpected, but interesting effect is it looks like work from home has a pretty substantial boost to fertility. So we're working on this, you know, and there's some issues around cause and effect, but my best guess is allowing people to work from home a couple of days a week increases fertility rates by something like 5 % to 10%.

27:34So it's not massive, but that's still a lot. The thing that's surprising for me is take the current US government. It's on this bender to ban work from home. The federal government is trying to ban it everywhere. At the same time, it's trying to promote fertility and increase birth rates, and they just completely clash. So, I mean, one view for me is for countries that are interested in trying to boost growth rates and birth rates in particular, promoting work from home is like an amazingly appealing policy. It's free for the government, It actually works. You know, it keeps people happy. It doesn't seem to have any obvious negative effects on firms.

28:07So, you know, why not do it? And, you know, that's a question for the current Trump administration. I'm sorry. What is the mechanism that you are suggesting? Is it that, you know, in that time you would have been spending commuting, instead you're making babies? Or is it something about, you know, expected work-life balance? What's going on there? There's one mechanism, I'll sum it up with my friend, that said, you can't get pregnant by email. So yes, it could be, you know, look, you're at home together. It looks like the much bigger mechanism is it's just easier to look after kids if you're at home, you know, one or two days a week.

28:44You both have kids and it's just easier to take them to the dentist, to the doctor, to pick them up from school. And what we see in the data is the effect is pretty equal for both, you know, both parents. So the effect is kind of doubled if you have both of a couple working from home rather than one. And primarily, you see, it shows up in kind of 30 plus grads. So you can think of what it is. It's more people that maybe have one, possibly two kids thinking, do I have another? And if you can work from home two days a week, it's just a lot easier and it pushes them over the margin. Nick, this has been a real treat.

29:13Thank you so much for joining me. Samaya, thanks so much for having me on the show. That is all for this week. You have been listening to The Economics Show with Samaya Keynes. If you want more economic analysis in your life, then do come along to the FT Weekend Festival on the 5th of September in London. I'll be speaking to Ed Conway about the hidden trade forces shaping our lives. The festival will feature debates and interviews with several familiar faces and friends of the show, like Martin Wolf, Dan Wang, and Sarah O 'Connor. As a podcast listener, enjoy 10 % off your ticket with code FTPodcast.

29:52Register now at ft.com forward slash festival. This episode of The Economics Show was produced by Misha Frankel Duval with original music from Breen Turner and sound design from Sean McGarrity. The broadcast engineer was Andrew Georgiades. Edwin Lane was the senior producer. Flo Phillips is the FT's head of audio. I'm Samaya Keynes. Thanks for listening.

30:20For years, credit scores were generated with just a snapshot, a snapshot that left millions of people out of the picture. So VantageScore is modernizing the credit score. VantageScore uses 400 % more credit report and alternative data than legacy models, bringing more people into the credit economy, like 5 million more mortgage-eligible people. Banks win and francs win. Lenders win and so do Lindas. When you build credit scores the right way, everybody wins. Get started at VantageScore.com. Hey, we're Nick and Jack, hosts of The Best One Yet. Behind every episode, there's research, email, spreadsheets, and our shared Google Doc, the Headline Hammer.

30:58With Google Workspace with Gemini, Google's AI is built right into the premium versions of tools we already use, like Gmail, Docs, and Sheets. No new tools or extra logins, just personalized AI for the way we work. That means less time on the work, around the work, and more time finding the stories worth talking about. Get AI tools for work. Upgrade to Google Workspace with Gemini at goo.gle slash workspace with Gemini. That's G-O-O dot G-L-E slash workspace with Gemini.