LinkedIn's CEO On How AI Changes Your Job — With Ryan Roslansky

0:00The CEO of LinkedIn is here with us to speak about how AI will change our jobs, whether we'll keep them, and where AI belongs in products. That's coming up right after this. Welcome to Big Technology Podcast, a show for cool-headed, nuanced conversation of the tech world and beyond. We're here today with Ryan Roslansky. He's the CEO of LinkedIn, and he's also a fellow podcaster in the LinkedIn Podcast Network, which we're proud to be a part of here at Big Technology. And you can find his show, The Path, in your podcast app of choice. Ryan, so great to see you, and welcome to the show. Wow, thanks, Alex, and thanks for that plug there on The Path.

0:34Appreciate it, man. I didn't expect that. Definitely. Look, as a fellow podcaster, I know if I'm appearing on a show, I want to make sure the audience knows right off the bat. And so it's the path. You can get it in your app of choice. There's some great guests on there, including Satya Nadella and Andy Jassy. By the way, if you have any booking tips for us about how to get those guys on the show, we'd be happy to take them. If you work for Microsoft, it's pretty easy to, you know, get your boss Satya on the podcast. So that's the trick there. You know, I was scrolling through your feed and I looked at it and I was like, this Jassy interview is a coup given that like Amazon and Microsoft are definitely competitors on cloud.

1:10So it's a great interview. We'll get into the podcast network in the second half, which we're proud to be a part of, as I mentioned. But first, let's talk a little bit about AI. You mentioned LinkedIn is a Microsoft subsidiary. LinkedIn hasn't been shy about building AI into the product. And also from your position as CEO of LinkedIn, you probably have one of the best views in the business world of how AI is going to change jobs, potentially replace them, change the skills that we have. And I think that's of interest to a lot of the listeners out there in terms of like, how's our job going to change?

1:42Or is it going to change at all? Will we be unemployed? Will we get paid better? We're not going to get into all of that, but I do have some questions for you. And let's start with your job. So there was a story in the New York Times in May 2024. So just a few months ago. It says, if AI can do your job, maybe it also can replace your CEO. So let's just start off with that one. Do you see that happening? I'd love to hear you weigh in on that front. Some days that would be very nice, actually. I would love that if someone else could do the difficult parts of this job. You know, maybe kind of going back to where you started, very fortunate that LinkedIn is part of Microsoft, I think a lot of times people don't even realize that.

2:26And to a certain extent, that's strategically set up that way by the acquisition that Microsoft did of LinkedIn roughly seven years ago. When Satya bought LinkedIn, the thesis was we can help LinkedIn grow by keeping it as a standalone company inside of the Microsoft umbrella. And to a large extent, that thesis has been playing out well When we were acquired seven years ago, we were, you know, less than a billion members and doing roughly$3 billion a year in revenue. Today, we're north of a billion members. There are seven members that are growing, joining LinkedIn every second right now. And, you know, we're doing north of$16 billion a year in revenue.

3:05So it's a nice place for LinkedIn to be inside of this Microsoft ecosystem. But also, you know, because every Friday I attend Satya Nadella staff meeting, which is just a master class on, you know, AI, where AI is going. and the future of technology in general. So very fortunate to kind of see what that looks like. And one of the things that I've learned or thought about navigating LinkedIn, where people are visiting LinkedIn every day to figure out how to navigate their career and what are jobs of the future going to look like, et cetera, is in a sea of uncertainty to kind of break it down in a way that at least can help people make sense of what's going on.

3:47And that framing that I use a lot of is that, you know, everyone's job in the world, your job, Alex, my job, everyone's job, at the end of the day, is a set of tasks that need to be done or skills that need to be learned to do those tasks. And if you start to break your job down into not just the title, but the set of tasks that need to be done on a daily basis, and then you apply the idea that, hey, there are certain tasks where AI can help better automate those tasks, you soon start to realize what your job can look like in the future. Now, if your job is just a set of tasks that can be automated, you need to start thinking about a new job.

4:29And I think in a lot of cases, you know, there's certain, you know, industries, sometimes, you know, translation or copywriting where it's like, huh, you know, AI is getting pretty good at doing some of those things. So what does that mean for my job moving forward? You know, there's a lot of, you know, jobs where AI is not really going to automate a lot of it, to be honest. And so, you know, those jobs are kind of on the other side of the spectrum. The majority of the jobs, though, are kind of in the middle, where a lot of the tasks that you do or a lot of the skills that you have are probably going to change to help you do your job better in the future.

5:04And it's not a new thing, per se. I mean, you know, one of the great things about LinkedIn is that all the data that we have on LinkedIn allows us to see remarkable insights and data trends over time. And if you go back to 2015 to now, which is just nine years, the average job on LinkedIn and the skills that are required to do that job, on average, have changed by 25 percent. So my job, your job, on average, those skills have changed 25 percent to do that job. Our guess is that, you know, it'll probably change 70 percent by 2030, like through AI. But I think that this, you know, change is a good thing.

5:41I think that it's going to give a lot of, you know, autonomy to people to do, you know, better work and to do better things. And sometimes people like to talk about, hey, you know, you know, these technology paradigm shifts or revolution end up in these really amazing, you know, breakthroughs or innovation or growth. And I think that's true. I think it's also important to realize that when you're in the messy middle of it, it can also be confusing and uncertain and people are going to be, you know, losing jobs and changing jobs frequently. And we hope, or at least I hope, that LinkedIn can act, you know, somewhat as a guide to help people navigate what that change is going to look like by helping you understand what skills you need to learn, what jobs are out there, how potentially, you know, your job may change and what we can do to help, etc.

6:28So that's where a lot of my focus is today. Right. And, you know, it's interesting because in the headlines, you read the headlines, one company goes out there and they say they're going to automate jobs with AI. And then all of a sudden it's sort of taken for granted that AI is coming for jobs. And in some cases, the company doesn't actually do it. But in other cases, they do. We had Sebastian Simeon Kowski, who's the CEO of Klarna on the show recently. And up until I interviewed him, I was like, there's no way AI is automating any jobs. And he convinced me they've taken AI and used it to automate about 700 customer service jobs within Klarna.

7:05But the thing is, you know, that doesn't seem to me to be the case across the board. My guess would be that that is more of the exception and not the rule. And you're the CEO of LinkedIn. You're sitting on top of like some of the most interesting data about how this technology is changing our workforce. So I want to turn it over to you. Have you seen companies either dramatically reducing their workforces or stopping to hire completely? And then you see a lot of AI in the job titles all of a sudden. Is that something that you've seen within the platform? The data we see so far, at least that I see real time right now, not seeing the former, not seeing companies reduce their workforce because of AI.

7:51I've definitely seen a ton of job descriptions that now include some form of AI in them. And at least I can speak for myself at LinkedIn. You know, one of the great, you know, we have a large customer service team, for example. And, you know, with a billion members on LinkedIn, we run a social platform on one hand. Then we run a set of these enterprise businesses. and so customer service at LinkedIn is a pretty diverse set of things where on one hand you've got a ton of people on a daily basis you know you know messaging in or emailing in hey how do I change my profile photo or how do I change my password or my email address and then you've got you know customers emailing in that are like hey we want to add a you know another thousand recruiter receipts to our, you know, contract?

8:42Like, how do we do that? And what we found, at least at LinkedIn, is we've been able to, you know, through virtual chat assistance, really help to automate a lot of those queries that really are not a good use of a human being's time to answer. So, for example, like, hey, I need to know how to change my password. Like, I don't think a human being needs to walk you through that process when, you know, an AI chat can really help you step by step, go ahead and do that. So, you know, on the flip side, when someone needs to, you know, add a thousand seats to the recruiter contract or has like a, you know, a billing problem they need to solve, et cetera, where to a large extent, success in solving that problem is not only, you know, good for the problem being solved, but also for the longer term retention of customers to LinkedIn, that's a much better use of a human being's time.

9:36So in our case, we have rolled out pretty aggressively chat assistance and customer care. And we've also increased the number of customer service reps at LinkedIn because it turns out that the more that our customer service reps are dealing with those kind of meaty problems, we're seeing better retention of our customers, lower time to resolution of those like thicker and more difficult problems and more success. So at least for us, we're seeing like AI doing a lot of great things and actually a need, you know, for more humans that are doing some of the things as well. So, you know, my guess is that over time, that's probably a lot of the direction where a lot of this goes.

10:20And I mean, look, I, you know, I probably like you am sick of hearing like this printing press comparison, but I think that there's something really important about thinking about technological paradigm shifts of the past and not only what happened with that technology, but what happened with the ecosystems around it. So if you were to go back to the printing press, wow, like here's this new invention. You look at it as an invention itself. It's like, it's pretty cool. Like, you know, the world can change and like you can do something differently now. And like, you know, the older way of doing it is going to change and people are upset.

10:52But more importantly, the printing press quickly spawned this huge ecosystem around it. So think about it. You need to have foundry for movable type. You need to have paper. You have to have ink. And out of that emerges newsletters or newspapers or books, booksellers that emerge out of that. And ultimately, it's all for the benefit of writers and readers who create new careers on top of it. And that quote unquote stack, if you think about that stack of things, the tech stack we talked about for the printing press, that translates super directly into a brand new economy. New kinds of jobs, new types of companies were created, the Enlightenment, democracy.

11:37All these things happen because of this technology and what spawned from it. So in my view, I think that's a lot of what's going to happen over time with AI, which is that you get into this messy middle we're in right now, and it's tough and confusing. But you think about it. Like you look what's happening right now around data centers or power or AI councils or responsible AI jobs. All these new things are starting to kind of, you know, pick up around this ecosystem. And my guess is that over time, an entirely new, extraordinary economy is going to be built out of this AI wave. Okay, Ryan. So outside of Microsoft Office, what personal uses of AI have you found helpful to you?

12:20Like, I want to know, like, are you a weekend speak with chat GPT type of person? Are you dumping some of the files that you have into some of these chat bots and saying, hey, I'm trying to lead this company. What am I missing? Or where do you find value from the AI tools as the CEO of LinkedIn? So if you take my job down to a set of tasks, you know, one of the most important tasks that I have is to communicate, to communicate strategy, to communicate, you know, direction of the company, to communicate vision, to communicate one-on-one, et cetera. And whether I like it or not, when you're in a job like this, your words carry a lot of weight.

13:01So you have to be really thoughtful, you know, what you say, how you say it, when you say it, consistency, et cetera. And I'm going to absolutely cheat on your question because I do find a ton of value. And my day lives inside of the Microsoft Office ecosystem. So, for example, if I'm writing an email to my boss, Satya, you're darn right that I'm going to push the co-pilot button just to like, hey, any advice on this email? It's like, hey, Ryan, either watch your tone or this sentence doesn't make sense. And it's actually super helpful to pull something like that together before sending an email off.

13:42Or another task in my job is constantly reading strategy documents or planning documents, et cetera. Just a lot going on for our company. We have 18 ,000 employees. We have five different business lines. There's a lot of complexity in it. So one of my hacks is inside of Microsoft Copilot is you take a document and it's like, hey, write me a quick FAQ on this document. and it'll take like a large document and you're like, hey, Ryan, here's like the really important stuff that you have to know about. And I found that to be really, really valuable. So, you know, at least for me, like the tasks of communication or just like insights, you know, kind of, you know, condensing information, finding the signal from the noise, I think have been super, super valuable, at least for how I've been using it on a daily basis.

14:33Actually, how about you, Alex? Remember, I'm a podcast host too, so I can ask you a question back. Like, how do you use that? Definitely. Well, it's interesting that you mentioned that you're like getting it to check your tone and see if there are examples that you mentioned and also just be like, can you condense this large amount of information down? And you started the question saying the answer saying communication is the most important thing that you do. And I was like, I wonder if Ryan's actually letting the AI write the things that he's communicating. and to me that's always sort of dangerous because ai will sort of bring your writing up to maybe the average maybe a bit above average if you prompt well and i think you want to be above average if you're the ceo of a company like you are so 100 it's interesting that like you talk about using ai effectively to hone your communication to check your communication almost as like this sort of good angel over the shoulder saying hey this is kind of the way where i would go with this or where I might.

15:33And that's almost exactly what I use it for as well. So I'll dump articles that I'm about to publish in and say, what did I miss? Or this is one of my favorite examples is I'll take the transcript of podcasts. Like I'll take the transcript of this podcast afterwards when we're done because you can pull it right out of Riverside and I'll put it in, I'll just say it, I'll put it in Claude. Claude is what I use for this, you know, maybe chat GPT one day and I'll be like rate this podcast on a variety of metrics and it will like look into flow it will look into difficulty of questions without me like giving the parameters and like go through all the relevant type of questions and be like all right this is how are the different topics and be like this is how you did with this and it's generally pretty good and then one other thing i'll do with it is i'll take like the um you know other interviews that i find interesting and if i want to like write about it, I'll put it into an AI chatbot and be like, I think this was the most interesting part.

16:34Why don't you call out some quotes? Right? So I'll never have it do the writing, but it will help me sift through the information and sort of modulate what I'm doing in a very effective way. I love that. And by the way, I think on the doing the writing for you, I both agree that is dangerous. And quite frankly, the technology is not quite there yet, in my view. For the most part, no matter what you ask, you know, be it Copilot or Claude or ChatGPT to do, to a large extent, it's going to, even with some intense, like, prompting, it's going to give, like, a pretty canned answer that you think it would be pretty much a lot of the same tone.

17:16I don't know, have a better personality that it would almost kind of pull back for anyone. You know, I think technologically over time, what's going to be fascinating, especially, you know, as we move into, you know, any of this agentic world is that personality layer is going to be a critical component of this, which is if I was ever to let something respond on my behalf, it would need to know so much about me and have the context and the background and the tone and the context of the people that I'm communicating with, my relationship with them to be able to do that. So it's not there yet. I think over time, that'll be a really important part of the stack, though, to figure out.

17:56It's something we spend a lot of time on because we build AI tools right now for recruiters, a large part of a recruiter's job. And their secret sauce, quite frankly, is when they see a candidate they want to reach out to, their own personalized way to reach out to that candidate. Some recruiters, for example, their secret sauce, like you're sending an important, you know, recruiting email to a, you know, amazing like Gen Z software engineer, like you include a lot of emojis, like that's the unlock, like that's how you get people's attention. So, you know, helping recruit, like helping the AI over time understand like, what would this specific recruiter do in this instance to help communicate with this specific candidate?

18:39Like that's a real unlock. And I do think over time, you know, the more we can help people be more effective in those parts of their jobs and own that personality with them, it will unlock and unleash a lot of people's ability to do more of the human parts of their job. So on the recruiting thing, for example, you know, a recruiter's job is to recruit. You talk to great recruiters, you know what they love to do. They love that stage in the process where they pick up the phone and they have a conversation, which is a blend of EQ, learning more about this person, potentially like understanding that candidate's tone and what they're saying as to whether or not they'd be a good fit with the group of people that they're going to be working with.

19:20And then potentially a lot of selling as well, because this is someone that's potentially in high demand. You want to help them to understand that this is the company for them to come and work for. And that unique blend of like EQ, understanding, selling is what recruiters love to do. What recruiters don't love to do is to sit there and sift through, you know, profile after profile after profile thousands of times, like find the signal from the noise, you know, write personalized in-mails to them to kind of get their attention. And so, you know, over time, we're trying to take a lot of that back into the product, into the technology that we're building that gives then those recruiters the ability to do the part of the job that they really love.

19:57And on the writing front, one more question for you. And maybe this is something that you've already considered building into the product, but the models are starting to give us larger and larger context windows. I think the Gemini one is 2 million tokens or some crazy number. And, you know, maybe there is a point where like you can take all of your writing, put it in the context window and say, write an email or an article the way that I do. Is that something that you're thinking about? Let's say for the recruiter that loves the emojis. Like, can we get to a point where you end up having LinkedIn say every member is going to have a customized AI trained on their communication style that will then help them be more efficient and communicating in their style to more people?

20:48So, yes, and even beyond that. So, by the way, that's a very thoughtful approach you just took there, which is that in your head, you're actually understanding how many tokens exist so that you can code into the prompt the right way to ask the question. I like that. My hope is that over time, though, it changes a little bit. So, LinkedIn has historically, you know, for the past 20 years we've been around, been the place where you maintain your professional identity online. So when I want to look someone up, when I want to look up Alex, like I go to LinkedIn and I see, you know, who you are, who you know, what you know, what you've done, et cetera.

21:29But you can think about over time LinkedIn also potentially being the place where a lot of that deeper information about you can be stored. So, you know, it's what I like to call, I don't have a better word for it yet, but your personality. You know, like how does Alex communicate? What does Alex tone in certain situations? Who are Alex's close contacts? Who does Alex have a closer relationship or more distant relationship with? You know, does Alex like to respond to things quickly or take time to respond to them? I mean, it's basically like your personality. And then if you can have that, you know, be it when you're writing an in-mail on LinkedIn or potentially when you're using like any other AI agents on the internet, the ability for that tool to hit that personality, to grab that context, and help it be smart in that response on your behalf, I think is a really interesting thing that we've just started scratching the surface on.

22:32We need to in the recruiting and the sales use cases because they're core components of who we are and what we do as a company. So for a recruiter, for example, It's not that every time a recruiter sends an in-mail to a candidate, they're going to have to, you know, input all their stuff and learn it on the fly. But a lot of these recruiters, we've been watching what they're doing for a decade. You know, how they respond, who they respond to, who they don't respond to, et cetera. And that, you know, quote-unquote personality will be built in to how they reply to candidates on LinkedIn. And we keep building that up.

23:06So, I mean, uncharted territory, exciting times. but again, for me, it all goes back to, you know, a recruiter's job is to recruit. A salesperson's job is to sell. Like all these other things that you have to do to get to that point. Like how can we help you be more efficient and effective in doing it? Even on the flip side, by the way. So not that you asked this question, but let me tell you about some other LinkedIn products. You know, if you're a job seeker right now on LinkedIn, which, you know, there's many that come to LinkedIn every day looking for a job and searching jobs. And you're staring at a job in front of you.

23:41And you're like, hey, is this job a good fit for me? You know, we've kind of created these AI tools in that flow, which, you know, you push the button. And it's like, hey, Ryan, like, turns out this job isn't great for you, because it requires these four skills that you don't have. But instead, come take a look at this job over here. Or, hey, Ryan, this job is perfect for you. I think you should apply to it. Great. Well, what do I do now? Well, you need to write a cover letter. Cool. Can you help me write a cover letter that like showcases the best parts of me and the best part of this job to give me like a template to start from?

24:12So, you know, we keep trying to build these like, you know, flows in the process to just help, you know, minimize the friction in any of these processes to help someone get to their end goal. Okay. And now I have a question for you about that because I hear efficiency and man, I'm pumped, but I'm also a little bit nervous because the other side of efficiency for the sender is volume for the algorithm for the receiver and you can get into a world where if it's so easy like like let's think about the job process part of this is proof of work like you write the cover letter a to introduce yourself but also to show that you're serious enough that you've written this letter that explains why you're the right candidate and that's why they should take a look at the resume and if they make people write cover letters, then they'll only get the more serious candidates.

25:00Now you can have an LLM just write us a cover letter for you and it's just as good as almost everybody else is out there. So how do you get around this potential threat of just having a world of AI slop, for lack of a better term, you know, both in the feeds on LinkedIn and on the inboxes where people are trying to find the information and that proof of work is actually important, right? Like if you have to spend time writing a LinkedIn post, you know, it better be good. Whereas like, if you can like prompt it, maybe it won't be. So how do you think about that? So a lot of ways. And I mean, it's a great question.

25:34Look, just to be like, you know, transparent and like upfront about how real this issue is. I think it was a couple weeks ago, you know, someone wrote an awesome, you know, bot on LinkedIn to like go and apply for like thousands of jobs, you know, as a bot, like doing this. And I think that, you know, the bot got actually a couple like interview requests back. And, you know, it was just it was it was really important and eye opening for us to understand, hey, this is something we have to be very thoughtful about, you know, this approach and make sure we have the safeguards in place. So your example is a real thing in the world.

26:10So there's a couple of ways that that we approach it. So number one, we've been investing really heavily on LinkedIn in what we call verified identity. And historically, you know, your LinkedIn profile, we didn't really have too much of a fake profile problem on LinkedIn because the barrier to create a LinkedIn profile is high. You know, you have to, you know, you have to create a resume along with your profile, basically. So, you know, a LinkedIn profile alone was just a pretty good sign of authenticity. But over time, we've even tried to push that a step further, given that, you know, through automation, you could theoretically find ways to do this stuff.

26:56Not to say we don't have a huge, you know, security team on the other side of this. But we introduced what's called verified identity, which is basically either through, you know, your work email address. If you are, you know, like active directory, if you're at a company using any of the office products or through your driver's license or through your passport, you can actually verify who you are. When you verify who you are, like a little check goes onto your profile, which is like, hey, this is really Alex. You know, we verified that this is actually Alex. So that's a great start for our product internally because, you know, when you're applying for jobs, when you are creating feed posts, when you are making connections, you know, real people, real verified people like are less likely to, you know, create havoc inside of the system.

27:49And it gives us much more flexibility in what those accounts have access to do versus non-verified accounts. you know, there are more restrictions on what they're able to do across the product. So, you know, a real verified account that's, you know, applying to a lot of jobs is most likely a real thing. A non-verified account that was just created, you know, five minutes ago, applying to a thousand jobs, you know, we'll stop that earlier on in the process. So one starts with just the identity of understanding who this person is in general, and then what they're able to do on the platform. Number two, you know, is to be very thoughtful in the middle of the matching processes that we have on LinkedIn about how we're connecting things.