Stewart Butterfield has been at the forefront of two epochal turning points for tech. First, he was the co-founder of the photo sharing site Flickr, that was one of the defining brands of the so-called Web 2.0 and the world of user-generated content. (It was later sold to Yahoo.) Several years after that, he co-founded Slack, which sold to Salesforce, and one of the big winners of the software-as-a-service wars, changing how people work and how companies operate. Now we're at another turning point for the tech industry. Layoffs have occurred across the space and AI is putting traditional business models into doubt. On this episode, we speak with Butterfield about his experiences and what he sees coming next for tech. This transcript has been lightly edited for clarity.
Key insights from the pod:
Low interest rates as tech “rocket fuel” — 03:35
How Slack changed the world — 7:04
AI and the innovator’s dilemma — 11:29
What will AI disrupt? -- 15:25
What does Salesforce do? — 17:49
Will AI change how software is built? — 26:44
Banking before computers — 27:40
Where regulation is still holding back tech development — 31:27
How Slack changed how workers work — 36:20
How do you measure software productivity? — 38:32
MMT and the trillion dollar coin — 40:07
---
Joe Weisenthal: (00:10)
Hello and welcome to another episode of the Odd Lots podcast. I'm Joe Weisenthal.
Tracy: (00:15)
I'm Tracy Alloway.
Joe: (00:16)
Tracy, you know, we've had a number of episodes about tech and some of the excesses of the VC boom, etc. Maybe the softening labor market, but it's one of these topics that we can't talk about enough because it's so critical, I think, to the markets, to the economy. And so, I think, ambiguous about where the industry is going next.
Tracy: (00:37)
No, absolutely. So there's two big things that seem to be going on right now, or two big questions. And the first one has to be, we've seen a lot of layoffs in the tech sector. And I guess the question is, is that reflective of the broader economy or is that something tech specific, as a more interest rate-sensitive part of the economy? And then I guess the second thing is, you know, tech was one of the big winners of the pandemic era. And how much of that is getting rolled back now? How much of, you mentioned excesses, how much of the excesses are being taken out of the system?
Joe: (01:11)
And then I would say the third thing, which is that, and it's just really leaped to the public consciousness in the last several months, is people being very aware of AI. And these questions about like, okay, we had this sort of era of what we thought of as a software company over the last decade. These “software as a service companies, VC funded, etc. So in addition to the macro questions about hiring, in addition to the question about financing, there's this question, well, “what's going to happen with tech business models?” And how upended [those] will be and what that's going to mean for existing talent and where the money is going to flow from all that. And I think, you know, everyone has posted some threads on that, but I've never found something that's like, “oh, that's that compelling.”
Tracy: (01:51)
You're right. There's a lot to talk about.
Joe: (01:53)
So I'm very excited because today we're going to be speaking with a guest that I would say has really been, I was going to say at the forefront of two sort of major epochal tech trends, but I kind of actually think it's two and a half because, so there's the Web 2.0, sort of social, kind of user generated internet. And then there was this sort of SaaS boom -- software as a service, etc. And then in the last few years with work from home, tech that enabled work from home, which is almost its own distinct kind of story in my view.
Tracy: (02:25)
Yeah. Someone who brings those mega trends together and maybe, maybe might be on the verge of another mega trend. We should ask.
Joe: (02:32)
Well, the perfect person to see if there's another mega trend. So we have the perfect guest. We are going to be speaking with one and only Stewart Butterfield, the co-founder of Flickr, the photo sharing site, the co-founder of Slack, the workplace messaging app that everyone used that eventually [got] sold to Salesforce, recently stepped down the CEO of Slack.
So Stewart, thank you so much for coming on. So I’ve got to say, I meant to say, what's also interesting here is both Flicker and Slack, as far as I understood it, originally started because you wanted to create an online game company. And both of them had to pivot into these like mega world changing hits. So I guess my first question is, are you working on a new online game?
Stewart Butterfield: (03:13)
I am not. I get ask that a lot and, and maybe.
Joe: (03:15)
It is a cliche question.
Stewart: (03:17)
It's a good warmup.
Joe: (03:18)
Yeah. Okay. But you're not right now.
Stewart: (3:19)
No, I'm not right now.
Tracy: (03:21)
So maybe we should start with the pandemic boom because, I mean, Slack really became the sort of poster child for the success of SAAS in 2020 and 2021. And you also had a kind of famous Twitter thread where you explained what it was like being on the front lines at that moment in time. But talk to us about how would you characterize that experience? Because looking back on it now, there is a lot of talk about maybe there was over-hiring, maybe there was excess in various ways, but how did you experience it?
Stewart: (03:54)
Yeah, that was a big moment, you know, March of 2020, but I feel like it really started 2015 or something like that. I was looking back through some old interviews and I think it was 2015, I might have the years a little bit off, Slack raised money in a $2.8 billion valuation. And people were like, “well that's crazy.” And even back then it was apparent and we would say in interviews, this is just zero interest rates. Like this is what happens
And so I think that kind of boost, that, I don't know, you could rocket fuel, but rocket makes it sound like... But I think this kind of like rocket fuel propelled us all the way through. And even, you know, before going back, I guess 2015, 2016, we started, we launched in 2014. We had a million dollars of ARR, you know, the following week. I think it took us about 18 months to get to a hundred million dollars of ARR. And that growth was crazy through that whole period. And then accelerated in March of 2020.
I think the excesses are maybe not a different story, but I guess there's two parallel tracks and one is like the Google-Facebook track, which is they make a lot of money and so there's no real constraint on hiring. And then there's the VC-backed company, which is they raised a lot of money and so there's no real constraint on hiring. But absent those constraints, you hire someone and the first thing that person wants to do is hire other people, right? Because it's a very obvious signal and it's very true that the more people who report to you, the higher your prestige, the more your power and the organization
Tracy: (05:24)
Empire building, right?
Stewart: (05:25)
Yeah, exactly. And even, you know, even if it's like a two-person empire, it's something. If, you know, if you become a manager, you want to become a senior manager. If you're a senior manager, you want to become a director. And it's a very powerful incentive. So if you hire a thousand people, you have, you know, 996 people or you know, roughly a hundred percent who are like, “I need to hire.”
So every budgeting process is “I really want to hire.” And that to me is the root of all the excess. Because if you don't have the constraint of “we just don't have the money,” you know, if you're manufacturing Lysine or something like that, you know, a 70-year-old industry where there's a lot of competitors and all the margin's been taken out of it. You can't do that. And if you have infinite money, either from being a monopoly on search engines or having VCs give you lots of money, you can get rid of that constraint altogether.
Joe: (06:15)
You know, you mentioned sort of this flood of VC money and maybe the sort of zero interest rates or cheap money was a sort of jet fuel or rocket fuel for it. But on the other hand, there was something real going on because the entire way the world consumed software massively changed probably between the great financial crisis and the pandemic. So how would you weight those two different factors? Like sure there's the cheap money and the VCs, but also there was this real shift in corporate spend on technology.
Stewart: (06:47)
Oh, absolutely. Sorry, I didn't mean to…
Joe: (06:49)
No, no, no. But I do think, you know, for listeners, it's easy to characterize the last decade sort of dismissively right? As like, well, cheap money, VCs, but something real really did change about how we used the the technology.
Stewart: (07:04)
Slack launched in the beginning of 2014. So not quite 10 years ago. Still, you know, it's like nine and a half years in and it set a $2 billion revenue run rate, you know, and we went public five and a quarter years after. We launched and, you know, were acquired by Salesforce a couple years after that. So that's very real. Not many companies in the world ever get to a billion dollars in revenue. And that's a real change in people's behavior. That's, you know, tens of millions of people who spend hours and hours a day in a fundamentally different way than they did before.
Tracy: (07:38)
So Slack launched in 2014, as you just mentioned. And since then, I think there's been a lot of competition from the big incumbents. Microsoft would be the obvious one, but I feel like there's always sort of mixed feelings in Silicon Valley about startups versus incumbents. Because on the one hand you think of a startup, maybe they're more innovative, maybe they're more flexible in what they can do, you know, they can zig when others are zagging. But then you do have these giants that just have lots of money, lots of capital, lots of talented engineers that they can throw at these problems and either acquire you, or maybe do a similar thing with their own resources.
Stewart: (08:18)
Yeah. It's funny, I used to go back to Wikipedia's kind of registry of what the components of the Dow were. And you can look, like a hundred years ago today it was American Leather Corporation and, I don't know, US Steel. Maybe it wasn't quite, but like the first tech companies were Westinghouse and General Electric and a couple
Joe: (8.38)
Amalgamated Jute.
Stewart: (08:39)
Yeah. And everything has a life cycle, roughly. You know, there's a couple companies that that survive for a hundred plus years. But all of the, and I was, now, we should all say the word H-E-G-E-M-O-N-Y -- at the same time, so we can see what the real pronunciation is, but…
Joe: (08:58)
1, 2, 3…
Everyone:
Hegemony
Tracy: (09:01)
We actually did a whole episode about the US dollar recently, where we said this quite a lot.
Stewart: (09:09)
That eventually disappears and I don't know how real the AI threat to Google is, for example, but Google has seemed absolutely invincible for a decade plus. It's inconceivable to me that anyone would compete with them on their ground, like to build that technology and to build the infrastructure and to build the data centers and stuff like that seemed impossible.
But, you know, maybe there's a way around, and obviously if you're a venture capitalist, you know, your job isn't to get a 15% return by investing in a basket of public tech companies. So you're very much incentivized to see the startups succeed. But they're also, you look back over the last 20 years, or even just make it from the great financial crisis, the last 15 years, it's been a lot of value created. I mean, not just in, you know, funny money, but a lot of revenue, a lot of shift in how business is done and a lot of huge successes.
Joe: (10:05)
Well, you know, I'm sure we're going to talk more about AI and how that's going to change the whole world, but it is interesting to me that the one big publicly traded company that a lot people are giving a lot of credit to for AI is Microsoft, which was like the dominant tech company 30 years ago, 35 years ago, probably. Does that surprise you? That like the biggest powerful juggernaut… And then of course, obviously, with the [Microsoft] Teams versus Slack competition, just that in this industry that's supposedly so disruptive and so competitive that there's like this company and a couple of companies that just seem rock solid.
Stewart: (10:41)
I think that that one, it's interesting. Very effectively done. But the AI tech is all open AI.
Joe: (10:50)
Sure. But they're like ahead of it and they did make that investment and people are excited … and all that stuff. So it's like tech aside, people are giving the company a lot of credit for how they're positioned in this
Stewart: (11:02)
Yeah. It's using their enormous resources, you know, huge profitability as leverage to leapfrog into the next thing. And I think it is very impressive.
Tracy: (11:13)
This was sort of why I asked that question about, you know, Microsoft versus startups. But what are the pros and cons of a smaller tech company versus a bigger tech company and, you know, you can shade it towards AI or talk about tech in general.
Stewart: (11:29)
Well, I mean, I think it's just classic innovators dilemma. You know, when you have nothing to lose, then you can do take any risks you want. And if you have a lot to lose, then you become much, much more conservative. I mean, I think Google and AI is a great example. Because obviously if you asked anyone two years ago, who's the greatest company when it comes to AI, everyone would've said Google, you know? 0% of people would've picked Microsoft at that time.
And why aren't they in Microsoft's position today? I think a lot of it has to do with that conservatism because, you know, they see we can release this thing and then it says dumb racist stuff, and then we get in trouble or we release this thing and someone relies on it and then someone gets injured or killed and an accident.
Or you can come up with a million reasons, you know. As I'm talking to the CEO of another company once and we're just complaining, we're like, “oh, our life is so hard.” But he was saying they're at a board meeting and the board said, “you really need to hire someone doing risk and compliance.” And he's like, “okay.” So hires that person and then the first thing that person does, like it started earlier was they need to hire two more people because they need, you know, there's a lot of work to do. We haven't, you haven't had anyone in risk. We've got to catch up and we've got to make sure we have the right policies and stuff like that.
Fast forward 18 months and they're doing the agenda for this board meeting and someone tells him that they're going to have eight people from the risk team show up and he's like “eight people? How many people work in that group? 23.” So this is like, you know, over the course of 18 months and every single one of, you know, it's just like the classic incentives. They get zero upside if they say yes to something and they have this powerful incentive to say “no.”
Tracy: (13:06)
Infinite downside, basically if they mess something up.
Stewart: (13:08)
Yeah, exactly.
Joe: (13:08)
Well, beyond just risk, however, thinking about Google and AI, it is, you know, AI inference, I guess it’s called, is costly, right? In a way that a search is not. People seem to, you kind of have to pay for it. ChatGPT is a service that people pay for. Google has been free, this whole [time[, you know, has always been free. Seems like some of the AI tools may not be as conducive to an ad sales model since the answer is right there rather than taking you somewhere else.
And so I'm curious, this tension that exists in Google where maybe AI is kind of like a -- I hate to use the word “paradigm” because I don't know what it means -- but a different business model, like the business of AI is than say the business model of search. All these companies, will many companies, whether large or small, sort of find that in a more AI-driven world, there's a real business model shift, even sitting aside some of the risks associated with it right now?
Stewart: (14:08)
A hundred percent. And I don't know that either I'm smart enough or I've thought about it long enough to be able to predict all the way out, obviously, but something like, you know, image licensing for stock photos. That seems like a business that I wouldn't invest in today when you can just go to Midjourney or whatever, they’ll get better and better.
Joe: (14:25)
I think some big stock photo site just had a buyout, or a buyout offer for a few billion. I don't know what's going on. Anyway.
Stewart: (14:33)
Oh yeah. Yeah. There's a funny story of a guy that said, “if someone gives me $4 billion, then I buy Getty Images.”
Joe: (14:40)
I will do that too if someone gives me $4 billion.
Stewart: (14:43)
But yeah, so there's like a pretty trivial example, but I think the longer term it's a little bit harder to know the knock on effects. Because you think about previous ways of technological innovation and I don't mean software computers, I mean just, you know, going back to steam power and stuff like that.
It would've been tough in, in 1840 to kind of predict the impact of railroads on the world. And then the automobile and all those technologies. But they really become part of us because you, when you think about, I used to use this example at Slack all the time. If your job is to dig ditches, you can only dig so many ditches a day. And if you are given a backhoe you can dig, I don’t know, a hundred times more, 500 times more, whatever the multiple is.
But those technologies come with, I think there's a better word than risks, but they come with additional risks. You can accidentally knock down a building with a backhoe and you can't accidentally knock down a building with a shovel. We get to the point now where there's this incredible augmentation to our memory, to our ability to think, you know, I remember going to the library when I was a kid and having to talk to someone, then they can maybe find the book, maybe they can't find the book. And now it's like anything I want know.
Tracy: (15:52)
The little card catalogues?
Stewart: (15:53)
Yeah, exactly. Now I can just like type in a couple words. I don't even have to spell the words right. I just have to like kind of gesture at the words that I want and Google will return everything I could want. What does that do to us long-run? And I think, you know, we're living with the effects now and this is pretty well recognized of what social media does and the doomscrolling and the, you know, changes to our physical posture because we're looking down at our phone all the time.
And I would really compare this to, we have a couple hundred thousand years of evolutionary pressure to seek calories and then now today, we live in a world of effectively infinite free calories for everyone. And so a lot of people get diabetes and we have a couple hundred thousand years of evolutionary pressure towards acknowledgement and recognition and all these social signals that suddenly you can get 1000X what you used to be able to get and you end up with a kind of cognitive diabetes.
So when I say the long-term effects are hard to recognize both in AI and just in what technology does to us as human beings in general, it can take a while for us to catch up. Like the Thames and the Charles River used to catch on fire routinely in the end of the 19th century. And we kind of figured out how to have benefits of the industrial evolution without those negative consequences. I think we'll do that with all kinds of tech, but it might take a generation or two.
Tracy: (17:21)
Just on the business model idea. Can you talk to us about how tech companies actually make money off of AI? And here is where I, you know, very embarrassingly confess that I still don't really understand what Salesforce does? And how they make money?
Joe: (17:37)
<Laughs> This is the big question we're going to have to, let's segment this for TikTok because a bunch of people have this question and no one has ever had the chance to get it answered. So we're going to have to do a little like online segment of Stewart's answer to this.
Tracy: (17:49)
Okay. So I guess my question is, one, what does Salesforce actually do? How do they make money? And then secondly, you know, if a company like Salesforce were to build something like OpenAI, how would they actually derive money from that product?
Stewart: (18:02)
Yeah, that's a good question. And I think that Salesforce is actually really well positioned to take advantage. Because here's what it does. I mean, obviously it makes money by selling software to people and it was the first to say, rather than just pay us a hundred dollars and you can have the software and run it on your computer like we used to do with, you know, floppy disks and stuff, we're going to sell it to you for much cheaper, but you’ve got to pay us every month or every year.
But what the software does, I think, is interesting because the original Salesforce product was CRM or customer relationship management and it's just, it's a database, right? It has like, here's what my customer is, here's, you know, depending on what's important to me, their phone number, or if I'm a dentist the last time they had a checkup. Or if I'm a retailer, the, you know, the total of all their purchases and you can kind of extrapolate from there to more and more sophisticated products.
So being able to do marketing segmentation and send the right promotional email to just the right people. And that's incredibly valuable to people. It's hard to build. It's expensive. And so Salesforce has a lot of customers that kind of rely on it for this core set of services that they branch out to a lot of things but are roughly around customer relationship management and marketing. So what do people do with that software? Well they, you know, they sit in conference rooms and they show each other slide decks saying “this would be a great promotional idea.”
Tracy: (19:19)
This is very relatable for me by the way.
Stewart: (19:25)
Yeah, I can't remember who said this, but someone's description of here's what a lot of people's jobs are and this is, this is actually a great opportunity for Slack. Their job is to get some data, put it into Excel, make a chart, take a screenshot of the chart, paste into a PowerPoint, and then email the PowerPoint to people. And then it really is, that's what 70% of people with desk jobs do-ish.
But, so, sorry. People are trying to decide how can we use this database of customer activity or marketing to make more to be more effective on our business. And that's almost certainly something that AI will do better or at least, you know, having an AI co-pilot alongside of you…
Tracy: (20:01)
So AI could, for instance, look at your proprietary database, I mean the software itself is not proprietary, but the data within it is unique to your company, and maybe spot opportunities within it that you as a human wouldn't have thought of. Something like that?
Stewart: (20:16)
Absolutely. I mean, so both people who are professional investors, I love the stories of, you know, they drive outside some parking lot at the mall and see which retailer has the most cars parked in…
Tracy: (20:30)
This is my favorite genre of sell side research, which it's when they send the analyst to like the shopping mall to look at foot traffic and stuff like that.
Stewart: (20:38)
But people do that inside their company too, right? They're looking for opportunities. Should we sell more of X? Should we sell more of Y? Should we sell more online? Should we, you know, open more smaller retail outlets so that customers can come in and see the product? Or should we concentrate on whatever? I think most of that stuff is you're going to do a better job with some companion, let's say, that finds those patterns more easily because they can just – “they,” the AI…
Tracy: (21:03)
I have this image of Clippy coming back and being like, “have you considered… ?”
Stewart: (21:08)
They can think about things much faster than you.
Joe: (21:10)
Well, I mean, the other area and just as you said, customer relationship management, it seems like AI would be really great at remembering birthdays. And for a salesperson specifically, it's like, “Hey, how did you know, oh, your daughter turned 16. You know, did she need driver's insurance?” Or something like that. And then it's like, well, does the salesperson even need to pay attention or can they just put in a rule that says every time I have a client that, you know, has someone in their family that has a major life event and just send them a message and something that seems like it's coming from me.
And so will it be good enough so that salespeople can talk to a hundred clients in a day versus 10, and not maybe like two out of those a hundred times say something so embarrassing that it destroys their entire reputation?
Stewart: (22:01)
Well, I don't know if it's good news, well whatever, the fact is that people who have done that for a while, you don't even need AI to kind of automatically get a prompt. You're sitting at your desk and it pops up and says, “you should email Customer X.”
Joe: (22:13)
Or you get happy birthday from the dentist.
Tracy: (22:16)
Joe, I’m pretty sure it’s not the dentist personally emailing you.
Stewart: (22:20)
But also I get a lot of, and like the dumb version of this is SDR -- sales development representatives -- or kind of people who trying to drum up some business for a company, send a lot of email. I think I probably get a disproportionate amount of this because I’m in a bunch of databases as a CEO of a company who will buy software from you. So I just get all of these ridiculous pitches like, “Hello Stewart. I couldn't help but notice that you are the CEO of Slack Technologies Incorporated…” And, you know, like something that a human would never write.
Joe: (22:46)
We get that too. People think that we can make purchasing decisions on behalf of Bloomberg all the time.
Stewart: (22:51)
So I think better versions of that, but I think even more, remember Google demoed that voice assistant that would call and make an appointment at the hairdresser for you. And it even like it paused and said stuff like, it was very convincing this image of like, I'm a salesperson and you know, you're a buyer at some company and my AI automatically generates this long email to you, and then your AI reads the email and summarizes it and just says, “Stewart has widgets to sell.” And then, you know, your AI generates this long response to me, and then my AI reads it. I can imagine a lot of activity just becoming like AI.
Tracy: (23:29)
So this was going to lead into my next question, and I'm trying to think about how to phrase this, but you can imagine how AI would be very, very relevant to an application like Slack. But I guess the question is how extreme do you think companies will go here? Because you know, there's a big difference between having a companion, as you put it, who is helpfully pointing stuff out and maybe, you know, going over your data and spotting things that you wouldn't have seen otherwise.
And then at the other end of the spectrum, you can also imagine where you have, you know, someone who actually uses AI basically to do their whole job. You know, you could just automate responses to Slack and retreat completely from that communications platform. So how far do you think companies will go with this tech?
Stewart: (24:20)
Yeah, I think they'll go all the way. And I don't know what all the way is, by the way, but I used to see this as an example all the time. Ben Evans, who is a former VC, I guess [at] Andresen Horowitz, wrote this great article called Office Messaging and Verbs. And in the beginning he takes these stills from the movie The Apartment -- Jack Lemmon and Shirley McClain, it's 1960. And Jack Lemmon works at an insurance company. And there's like all these, you know, so the early in the days of office buildings, all these shots of like long rows and columns of desks and on his desk there's an adding machine, there's a typewriter, there's a telephone, and then people come by with a push carts, you know, with paper on it and they'll put some paper on his desk and then he'll perform some calculations and then type up the results and then put the paper on the other side of his desk in his outbox.
And someone will come along with a trolley and take that. He's literally a cell on a spreadsheet, that is exactly what he's doing. It's like take input, you know, execute formula, produce output. And that's what Ben talked about. Each floor of this insurance company's office is like one worksheet in a big spreadsheet. The same number of people work at insurance companies today. No one does that anymore. So when you say that people do their whole jobs with AI, that'll last for a little while. You know, let's say if you’re, you know, a content marketer and honestly you couldn't, AI couldn't do worse than most of the garbage content market..
Joe: (25:39)
Facts. Facts.
Stewart: (25:40)
But eventually people would say, look, that's not your job anymore. No one's job is to perform arithmetic anymore.
Tracy: (25:45)
That's a good way of looking at it. I just remembered my first job at Bloomberg when I was an intern was to monitor the fax machine and to actually like bring the important faxes to someone. And that job has thankfully gone away and I don't feel that bad about it.
Stewart: (25:59)
Over the last, I think hundred years the pace at which jobs become obsolete has definitely increased.
Joe: (26:04)
I mean this idea that, you know, tech will not end employment as we know it, it'll likely change and certain specific careers or categories will go away and new things will [be] created. But as someone obviously in your career who has hired many engineers and coders, and this is a whole ‘nother aspect of AI, is like people are kind of blown away frequently. How do you see that specific[ally] changing or will there be any change…
Tracy: (26:30)
Right, because coders specifically are freaked out, right?
Joe: (26:32)
And is there going to be an infinite demand for the skills that we call coding or engineering today? Or will what we think of as a coder sort of be a different skillset?
Stewart: (26:44)
It's a little bit harder to extrapolate, but definitely it will become different. But even people kind of tending to these machines who are generating code for them, software's really hard to make. And it's like surprisingly hard. And that's why there's so much crappy software. I don't know exactly what their supervisory duties will be, but it is going to be just like any wave of technology. It's this massive augmentation and then people end up moving higher up the value chain. Just like, you know, I sometimes think about banks in, I don't know, 1910s or something like that. I have no idea how they knew how much money anyone had.
Joe: (27:21)
I think about this all the time. How did…
Stewart: (27:23)
Talk about card catalogs? They’re like looking up…
Joe: (27:24)
I think about this all the time, how did stocks work before?
Tracy: (27:27)
You just opened the vaults and you see what’s there?
Joe: (27:29)
I'm glad you said this, it makes no sense to me that we had any banking and finance before computers. I just cannot wrap my head around it.
Tracy: (27:35)
Well, you also had a lot of bank failures, which suggests that they weren't very good at it.
Stewart: (27:40)
But then, imagine being like a stock trader in the 1960s, than in the eighties, and then in the two thousands and now, you know, like the 13 milliseconds you get by moving your servers closer to wherever in New Jersey make a real difference. No one was doing that before, but again, there's not fewer jobs. In fact there's probably a lot more jobs that are trying to make money off trading than there were in the sixties or the eighties.
So I think the same thing is true of coders, whatever that role evolves to be. You know, like no one in the sixties was hiring physics PhDs to be quants at some hedge fund. But now look, there's another pathway for people who do PhDs in physics. And so I think the same thing will be true of coding as well.
Tracy: (28:21)
Since this is now firmly an AI conversation, can we ask, you know, is this something that you're interested in and is this maybe what your next project could be?
Stewart: (28:32)
Weirdly, not really. I mean it's, I'm sorry, I'm interested in it as like a human, as a citizen. I haven't come up with anything where I'm uniquely able to contribute. I don't think in AI, maybe that'll change. Because you know, it'll become part of the background of everyone's roles and we’ll, you know, come to rely on it more and more. Just, you know, this is probably a crappy example, but word processing or, you know, I would never would've been someone who is letter press layouts and moving bits of lead around on something like that. But it definitely changed the way word processing changed the way I wrote because it's not longhand and I can infinitely edit it. And I wrote a thesis, like in grad school, and all that stuff.
But maybe a better idea or a better comparison is Excel. Because I think about in my life I might have done as much financial modeling as all of humanity did until 1965, something like that. Because when you can just make a spreadsheet and then be like, oh, change those, changes, change, everything cascades through. Whereas before that was, you know, like a lot of people with ledgers and calculators and pencils and paper and allows you to think at a much higher level. So I think about AI again in that augmentation capacity and what it'll enable us to do.
Joe: (30:04)
Just actually going sort of back to the last decade or two decades, I mean, you know, Slack was sort of like the, to my mind, one of like the prototypical like software as a service companies. But there were all these other ones and many of which we've never heard of that probably made a bunch of money by like, we're going to improve, you know, dentist billing or ticketing at sports events.
There's just all kinds of little niche categories that someone found an opportunity for. Is there still, I don't know if low hanging fruit is the right word, but moderate hanging fruit in that world that has not been exhausted of essentially still taking like Tech 1.0 and really legacy tech we've talked about a little bit with like Patrick Mackenzie and, you know, whether it's government, and just sort of like, is there still a lot of the economy that hasn't really even got to like 2010s tech yet?
Stewart: (30:53)
Yeah, I mean I think anytime you see you're in a financial transaction with someone and there's pen and paper involved there's an opportunity there. And I think a lot of the blockers tend to be regulatory. It's like when you go to the doctor and you're given like seven pieces of paper and you have to write your name on five the pieces of paper, sometimes you have to write your name two times on the same sheet of paper and the date over and over again. And that's a regulatory block.
Tracy: (31:15)
This is also my big gripe about FinTech, right? Which is you get a bunch of people going like, “oh, it's ridiculous that we're still writing checks nowadays” or doing this or doing that. But often the reason is a regulatory constraint versus a technological one.
Stewart: (31:27)
Yeah. So I think, eventually those regulatory dams break because well, maybe I shouldn't say with such confidence we hope. But I can't imagine that a hundred years from now, we're still getting seven pieces of paper when you go to the doctor's office and have to write our name on five of them. I think there's a lot of opportunities for real improvement and the low hanging fruit maybe never ends. Because as soon as you automate some layer and people start operating, you know, at a higher level, the people who are doing financial modeling aren't doing arithmetic anymore, they're thinking about the business, then there's new opportunities for automation that come at higher and higher levels.
And that's what we really thought about at Slack. The ability to -- this is maybe a little bit abstract -- we thought about Slack as like a messaging bus inside of a computer. It's the like interchange or the traffic controller for all the thoughts that people are having. And a lot of those are just like, “yeah, want to get lunch?” or something like that. And some of them are “here’s my extensive proposal for next quarter, blah, blah, blah.” When you're able to improve the efficacy of communication versus the I’ll send a paper memo that can schedule a meeting and it's three weeks from now, or something like that.
It opens up new possibilities. And the same thing is true with every bit of automation that you can do. There's a huge amount of business processes that are essentially humans translating between one database and another because the databases aren't connected effectively. And eventually those databases become connected effectively and people move up the chain again and get to work on harder stuff? Or I guess, you know, stuff that produces more value.
Tracy: (33:06)
AI to standardized databases would be amazing. Just on the topic of constraints, and going back to the first part of this conversation where you were talking about how, you know, a lot of the empire building or hiring boom that we've seen in tech companies was potentially driven by low interest rates and, you know, you didn't have a financial constraint on the amount of resources you could accumulate.
It does feel like in 2023, Silicon Valley, the tech industry in general, is in a very different financial environment. We have higher interest rates, we just saw Silicon Valley Bank fail and that seems to have taken a big chunk of potential liquidity out of the market. How do you expect that to impact the industry and what are you seeing now?
Stewart: (33:49)
Well, obviously we've seen a lot of layoffs and I think probably, you know I never want to suggest that being laid off is a good experience for anyone. For the companies, it's almost certainly a healthy thing because a lot of companies are, you know, that have 20,000 people could probably do what they're doing with 12,000 people or something like that.
The higher interest rates, I think, I don't know, I don't want to say that they're good, but you go back to 2009, which is when the company that ended up becoming Slack was founded. This is, you know, like March of 2009. So I just said we're still probably in the great financial crisis or, you know, maybe just getting out of it. And if you're a venture capitalist like Excel, you could buy 20% of what became Slack for $5 million. You know, and then you end up making $4 billion or something like that.
So I think from if you're a VC, it's probably not all bad. You know, some of your existing investments I think lost a lot of value and maybe you're going to have to defer realizing a lot of gains that you might have realized much more quickly a couple years ago. But over the long run, I don't think it's a net bad for them. For the companies, it can be a net bad.
Joe: (34:59)
Can I ask like a sort of cultural/economic question, which is like, I always thought, I think Slack is like an acronym for something, right?
Stewart: (35:08)
Yeah. It's unclear…
Joe: (35:12)
It’s unclear? You're the one who created it…
Stewart: (35:13)
We have a transcript of a conversation we were having internally where I suggested as “Searchable Log of All Communication and Knowledge,” I think. But I'm not sure if I if we came up with the name Slack first.
Tracy: (35:24)
Slack is catchier, I gotta say.
Joe: (35:26)
But the reason I actually asked the question is because the last decade, in addition to everything else, was characterized by labor market slack. And so I always think it's sort of interesting that this big company -- Slack – came [from] the decade of like loose labor markets. And it also seems that your technology changed in many ways the relationship between management and employees and employees and places can unionize more easily because they have Slack channels that they can communicate in ways that maybe the management doesn't love or maybe managers at companies want all their workers back to the office and it's like, “well yeah, but we can communicate on Slack, etc.” Do you have any sort of like broader reflections on the way, because this is not really a software question, it's a very Slack-specific question of how you saw firsthand your software changing this sort of employee-management [relationship] or just this sort of corporate environment?
Stewart: (36:20)
Yeah, I think going back to that idea of the ditch digger given a backhoe, most technology kind of amplifies our ability to accomplish something or augments us or gives us extra power. So that's true everywhere. Obviously ,where you intend it to be true, that's great. Where you don't intend it to be true, sometimes it's great, sometimes it's terrible, sometimes it's somewhere in between.
But the superpowering [of] people's ability to have conversations at work will include conversations that sometimes managers wish that they didn't have. I do think that's something that works itself out relatively quickly. And I also think, interestingly, internally we will call that Slacktivism, right? And probably someone else came up with that outside, but I think that's probably moderated quite a bit over the last year just because the environment has changed.
Because the other, you know, fact that comes with the low interest rates is that people pay a lot of money for engineers and engineers can always change jobs and get paid more and more money. And that's less true today. People are more worried about job security. So there's less, and I'm not saying this is a good thing necessarily, but there's certainly less labor organizing happening on Slack than there would've been two years ago.
Tracy: (37:30)
This is a slightly weird question maybe, but I mean, one of the debates about work from home is are people actually more productive in an office environment or at home within Slack? Did you ever have conversations about maybe measuring productivity in some way using the communications channel?
Stewart: (37:48)
We did and they were all, they all seemed kind of doomed. You know, I had a, I don't know if she's still the CIO, but a woman named Lori Beer was the CIO of JPMorgan, maybe still is. If so, hi Lori. And we had this great conversation once, we had breakfast at a conference and we were talking about the history of trying to measure the productivity of software engineers, which is famously impossible. So you would say, well how many lines of code do they produce? And then people would just change their coding style to be, you know..
Joe (38:15)
Verbose?
Stewart: (38:16)
Yeah, exactly. Or you would say, how many bugs get closed? And then instead of just fixing something right away, people take the time to go file the bug and then fix it and then close the bug.
Tracy: (38:24)
Insert a bunch of bugs and then you're the solution to your own problem.
Joe: (38:28)
Goodhart's law, right? As soon as you start to measure something, then it doesn't work.
Stewart: (38:32)
Yeah. And so you kind of, I don't know what we came up [with], or the way the conversation kind of ended was looking at employee engagement surveys. If engineers are happy, it doesn't necessarily mean that they're productive, but if they're unhappy … do I have the backwards?
Joe: (38:50)
No, no, that's sounds right. If they're unhappy, it probably means they're unproductive.
Stewart: (38:52)
Yeah, because no engineers are not productive and happy. It's frustrating to feel like you come into work and you put all this effort in and nothing happens. You know, no one wants to work on dead end projects that get canceled and all of that. So the same thing I think is true, if you think it's hard to measure the productivity of a software engineer, like marketing, it's impossible to measure.
People just have no idea whether their marketing department needs to be five times bigger or five times smaller or do they need one at all? I'm not sure what you do about that, the measurement, because there's a lot of just kind of hand feel and management it would be great if you could in some sense, but I think you also run the risk of eliminating all the opportunities for serendipity and accidental discoveries and cool innovation.
Joe: (39:44)
I have one last question and Tracy's going to hate me for asking it, I already asked the cliche question.
Tracy: (39:48)
I know what it is.
Joe: (39:50)
I already asked the cliched question about whether you're going to start another games company. Do you still support minting the coin to avoid a debt ceiling disaster?
Tracy: (39:57)
It’s unavoidable. We cannot have a single conversation without talking about the coin,
Joe: (40:02)
No, we can, but when there's a prominent famous person who on Twitter has expressed support, I mean…
Stewart: (40:07)
I feel like that's, you and me, we bonded over the coin. So yes, I think you definitely could. I think, you know, it's really funny, the debate about modern monetary theory seems to have come down to should we spend lots of money? And if you think “yes,” then MMT and if you think “no,” then you're against MMT, which doesn't seem to capture it at all to me. Like the idea that the deficit is a myth seems just obvious and that the government's role in the amount of money is like the adjudicator of a basketball game deciding how many points. You can't have too many points. I mean, you can mess up the game of basketball so it's not fun for anyone. Like, you can cause a lot of in inflation, but that idea that you need to have the money in order to be able to spend it is obviously not true if you're the government.
Tracy: (40:50)
For what it's worth, I'm not a big fan of MMT, but I do think the government should spend money. So there we go.
Joe: (40:55)
Stewart Butterfield, this is such a treat to have you on. Kind of funny that our first guest in our new physical video studio is the founder of Slack, which more than anyone else contributed to the ability to work and operate and communicate remotely. But such a treat to have you in. And thank you so much for joining us.
Stewart: (41:13)
Thank you.
Tracy: (41:14)
Next interview we can do via Slack, just to even it out
Joe: (41:31)
Tracy. I thought that was a lot of fun. That was a real treat. That was just very, I don't know, in my mind I had that hyped up and I felt like in my mind it lived up to my own hype.
Tracy: (41:40)
It was a wide ranging conversation and I love that we went from how did banks use to do business in the early 1900s to the future of AI and the labor market.
Joe: (41:51)
We're going to have to do an episode on just how a bank worked a hundred years ago. Because in my mind I do not understand how they kept track of ahow much money anyone had before digital technology. Like they really just what had a piece of paper?
Tracy: (42:04)
No, you just opened the vault and see what's in there. It’s like, yeah, we're all good. And we're all good today. We're all good. No, there was a lot to pick through in that conversation. I mean, I do think the idea of finance as being a constraint on resources and the workforce. I mean, it is kind of obvious, but also we are seeing it play out in real time right now. So it was good to hit that.
Joe: (42:25)
No, and just this idea, it does feel like there's multiple turning points at once and I don't feel like anyone knows. It's like pretty clear, right? It’s like, “oh, well this is who's going to win? This is who's going to make money. This is going to be the business model.” It does feel like even if we didn't have this sort macro change, there would be a tremendous amount of ambiguity about how value is going to accrue and how companies are going to work and all of that stuff.
Tracy: (42:53)
Well, also the conversation about Microsoft and how, you know, two years ago no one would've expected Microsoft to be the sort of front runner in the AI game. And yet here we are today. That was really interesting. Do you have any bets on what Stewart's next project is actually going to be?
Joe: (43:08)
Even though he said “no,” I still feel, well he's probably going to do a game.
Tracy: (43:13)
He cannot resist the lure of videogames?
Joe: (43:15)
It’s like, if you tried twice, you know, why not?
Tracy: (43:17)
Third time's a charm.
Joe: (43:19)
Yeah, that’s my bet. And then it'll pivot to something and be another world changing thing. That's my guess. Shall
Tracy: (43:23)
Shall we leave it there?
Joe: (43:24)
Let’s leave it there.
You can follow Stewart Butterfield on Twitter at @stewart.