Grammarly CEO Rahul Roy-Chowdhury: Trust, AI, and the future of work

0:00Scaling a business in the US can feel like a maze. Each state has its own payroll, benefits, and compliance rules, pulling your focus away from growth, which is why founders use Deal. Deal is the professional employer organization, or PEO, that gives you a dedicated HR expert, plus Fortune 500-level benefits for your team. The National Association of PEOs says businesses can grow twice as fast if they use one. So, if you're scaling, make it simple with Deal. Go to deel.com. slash MOS and get up to three months free. If you're in private capital, you understand that your next great deal starts with who you know.

0:41That's why over 3 ,000 firms trust Affinity, the CRM platform purpose-built for private capital. Affinity's relationship intelligence reveals your team's strongest paths to founders, investors, and decision makers using insights It's hidden in emails, calendars, and meetings. With powerful AI tools like Affinity NoteTaker and Deal Assist, you'll spend less time managing data and more time closing the right deals. Learn more at affinity.co slash marketplace.

1:18When I was in college, I was a math major. and I was interested to understand how we could develop a model of the world to start inferring facts and knowledge. I came across Alan Turing and a future he had imagined on intelligent machines. Alan Turing's work, this theoretical construct of what computation actually is and what it enables, was, I thought, just a deeply compelling and deeply creative piece of work. Long before Rahul Roy Chowdhury was the CEO of Grammarly, he was a PhD student obsessed with Alan Turing, considered a founding parent of computing, and with Turing's ideas about what we now call artificial intelligence.

2:11But Rahul wasn't content to sit around studying for very long. I published a paper in a pretty prestigious AI journal. It was kind of a big deal that I had this paper published pretty early on in my PhD career. But I felt that I wanted to have a bigger real-world impact. I was wondering how many people even read this paper. This is not a value judgment in any way, but I felt like I needed the real-world impact of the work that I did to really feel motivated to do that work. So I realized that maybe academics wasn't my path. Dropped out, joined industry. and here I am, decades later.

2:54Rahul spent most of his career leading technical teams on ambitious projects, like building the Chrome browser at Google. In 2021, he joined Grammarly as its head of product. Soon after, he took over as CEO. Rahul is tasked with steering the$13 billion company through a chapter of profound disruption. For most of its 15-year history, Grammarly's writing assistant software has been a market leader. But new competitors are crowding the field. Tools powered by large language models like ChatGPT, InflectionsPie, and Microsoft's Copilot now offer to improve your writing too, sometimes for free. I talked with Rahul about how Grammarly is handling the threats and opportunities that come from the latest AI advances.

3:41You've got to have incredible talent at every position. It's like this huge push. There are fires burning when you're going home. Can you believe it? Such an idiot. And then you go back to, this is totally going to be amazing. There are so many easy ways. I have no idea what to do. Sorry, we made a mistake. But you have to time it right. Oops. Working out of a three-bedroom apartment. Stuff that just seems absolutely nut balls. Ten years later, we're like, well, that's just how you do it. We haven't made it just how you do it.

4:12This is Masters of Scale.

4:20I'm Jeff Berman, your host. Before we jump in, just a note that Grammarly is a paid sponsor of Masters of Scale. Our chat with Rahul was planned, recorded, and edited separate from that relationship. Rahul, welcome to Masters of Scale. It's great to be here. Thanks for having me, Jeff. It's great to have you here. I was tempted to ask Grammarly to help me with my opening question for you, but it's a simple question. So for those who haven't worked with Grammarly, how do you describe what Grammarly is and what you do? Yeah, this is something Grammarly could have helped you with if you had asked.

5:02But the mission of Grammarly is to improve lives by improving communication. We're here to help people communicate better, to communicate more effectively, to get your point across, to be understood. Why do you think that's so important, particularly now as we sit here talking in 2024? Why is clear and effective communication important enough to be the core mission of a company? Well, communication has always been important. If you want to go way back, it's what defines us as a species. We're a social species, we communicate with each other, and our success depends on how well we can collaborate and make our intentions understood to a large group of people.

5:42What's made it particularly important is a couple of things. One is I can actually use software to help me in doing this important task. But then over the last couple of years, teams have become more distributed and more global. And through the pandemic, it accelerated what had already been something that is in progress, which is more remote work and more sort of distributed teams communicating increasingly asynchronously. And so as a result of all these changes, the weight of the written word was more than it had been in the past. And so the importance of using a tool like Grammarly to make sure that those written words are impactful and have the desired effect is more important now than it has been ever before.

6:27I'm curious, when you came in to take the CEO role at Grammarly, was there something from your personal life or your own work life that really animated this as a mission for you personally? The thing that animates me and defines the choices I've made in my career are I'm a tech optimist. And I want to work on the hardest, most impactful, most meaningful problems that technology can solve. So way back when, in the early days of my tenure at Google, there was a project to build a better web browser. It was called Chrome. When I joined Chrome, success was far from assured. It was this mission of improving the web platform.

7:14And I loved that mission. I loved the idea that we could create a platform that anyone anywhere in the world could access. It was an incredibly rewarding journey a decade in. Fast forward to 2021, I come across Grammarly, and I had that same feeling that I did in the early days of the Chrome browser and the web platform. And I felt like, you know, communicating well is such a hard but such an important problem to solve. And if technology can help you do it better, what a great mission to sign up for. About 18 months after Rahul joined Grammarly's head of product, the large language model revolution took hold.

7:55Opening AI released ChatGPT for public use in November 2022. That opened the floodgates for AI-assisted, well, pretty much everything. Grammarly knew it needed to adapt and quickly to this technological shift. So Rahul was asked to take on the role of CEO just a couple of years into his tenure. Two of three co-founders are still heavily involved with the company. With their support and his background in building tech products, Rahul is well positioned to lead. But the job can still be daunting. AI is not for the faint of heart. The pace of change in AI is unlike anything I have seen before. I've been thinking about AI for a while.

8:37Way back when, in fact, I was in a PhD program in AI. I dropped out. I didn't finish. But it's been on my mind. And I've never seen the level and speed at which the technology is evolving and the capabilities are evolving and getting better. That's a wonderful thing. I'm a tech optimist, and I think this is going to make all of our lives incredibly strong, powerful, successful. It's going to be this big way to augment what we do every single day. And there's ambiguity and uncertainty in the road ahead because the pace of change is so fast. But my experience is working on meaningful things is hard and ambiguous, and that's what makes it meaningful and worthwhile to pursue.

9:20We've done a number of master's of scale episodes with successor CEOs. Some of them are what are often referred to as re-founders. They're tasked with really almost reimagining the company for this next phase or actually doing so, as opposed to continuing on its current path or slightly evolving it. Do you think of yourself more as a refounder or more as an evolutionary CEO of the company? You know, I'm not sure that either of those really neatly fit the place you're in. How would you describe it? What I describe is, for example, we have a longstanding investment in user trust, privacy, security.

10:03We've really invested deeply in this area, and those investments are going to be paying massive dividends because user trust is a core value, but they're even more important now in the age of LLMs. all this data, what's happening to the data, is my data safe? What's the motivation of this company that's collecting my data? We've also been extremely focused on making sure that we are bringing technology to bear to solve real user problems. Seems like an obvious thing to say, but often you can end up getting enamored of the tech and lose sight of actually how it's going to help a user. And so we want to keep all the things that have made Grammarly what it is today, and then reimagine how we can think about increasing the value we provide to users in the age of AI.

10:51So it's not a refounding, it's not a evolution, it is taking what makes the company great and special and set our sights to the future. Grammarly has been at this for 15 years. You're the market leader. You can't listen to a podcast or put on a sporting event without hearing or seeing a Grammarly ad, But you're facing a new wave of competition that is rising about as rapidly as scaled competition has ever risen in any industry. So how do you think about that competition and maintaining your market position? I welcome it. I'm a firm believer that we want to stay focused on solving real user problems.

11:34That's always the foundation of everything we do. I don't want to spend a lot of my time looking over my shoulder at what other people are doing because that's time taken away from how can I uniquely provide a solution to that user problem. You can think of a simple view of the lifecycle of communication being there's a conception phase, there's a composition phase, there's a revision phase, and there's a comprehension phase at the other end. And so we have historically focused on primarily the revision phase of this lifecycle. Not because we wanted to, but because the technology didn't really enable us to help people along the full spectrum of this communication journey.

12:17Now with large language models, we can. We believe Grammarly can really reimagine how we communicate across all these different tools. I'm sending you an email, and maybe I want to include some context from a project update. I just had a conversation about that project update in a different channel, maybe in Salesforce or maybe in Slack. Now, Grammarly is in all of these different forums. And so maybe we can pull in the right context at the right time to help you do your job better. I think we can be more effective at work. And I think we can remove some of the drudgery and make work more fun.

12:54I'm sure we're not the only people who have seen this opportunity, hence competition. But that's great. That's how a free market works. And I think that's good for consumers. Have there been specific customer stories that have really motivated you or helped shape your product roadmap? There's a specific person I want to call out here. His name is Mark Francis Long. He's dyslexic and actually has a dyslexia advocacy group called I Am Lex. And he has talked really movingly about how Grammarly has helped him navigate his professional life. He sees Grammarly as a tool that is looking over his shoulder, helping him correct and remove all the inaccuracies in his communication that he felt were holding him back.

13:42And so he has this very inspiring vision of, that's the vision of I Am Lex, which is there's a superpower in dyslexics seeing things and pattern matching differently. And so how do you harness that superpower, but get the help you need? And that's the role that Grammarly plays. And he has a TED Talk. It's wonderful to, I would highly encourage everyone to see it. But that's one of those inspiring stories. And we have dozens of those every single week. That is what motivates me to come to work every day. I love that. And I mean, you know, the legion of dyslexic business leaders and political leaders, I mean, Richard Branson, Charles Schwab, Governor Gavin Newsom in California, who have that superpower, but also have had to work really hard to overcome the learning challenges.

14:27These new AI tools, including Grammarly, are just such a massive unlock for so many millions of people around the world who have those learning challenges. Two of my kids are dyslexic. And so, you know, tools like Grammarly are lifesavers. My kids are better students. They become better writers. And yet they go to a school, which is a wonderful school, but that at least for this past year banned the use of AI tools beyond effectively Google spellcheck because of concerns about, you know, cheating and, you know, where the lines are. And I have a lot of empathy for the school administrators. And yet I'm looking at this going, this is insane.

15:11These are the tools of the next decade plus, they need to know how to use them. They need to be literate in them. So given Grammarly's leadership position in the market, how do you think about educating not just companies, but schools about these tools and how to responsibly deploy them? Yeah, we are in active conversations with literally hundreds, maybe thousands of universities across the US, across the world, all grappling with this question of, there's this big technical shift. I'm afraid of it on the one hand, but I see the potential of it on the other hand. And also I feel some pressure to make sure my students graduating from my Institute of Higher Learning are prepared for the jobs that are going to be created in the future based on AI.

15:59So we, first of all, want to have a constructive dialogue. I understand the concern and the fear. It is very new. It is moving very, very quickly as well. And so we talk to educators on a regular basis. We learn from educators on how we could improve the product or change the product in ways that ease some of those concerns. And also just educate folks on this technology is not something to be feared. It's something to be embraced. And if pedagogical mechanisms need to evolve to take account of this new technology, that's great. As an example, one of the features we added to our product for education specifically, which I love, is you can use generative AI in Grammarly, and maybe you're writing an essay or doing some other piece of work product for your college, and then Grammarly at the end can offer to cite your use of AI.

16:55So what are the prompts you ask Grammarly for? How did you actually use AI? And so if someone, the only prompt was, give me a 500-word essay on the Civil War, you didn't engage in the material. I can see that. But if someone's prompt was, help me strengthen my second paragraph on this point, give me a critique on whether I'm actually persuasively using data to make this point. And then you can see that this interaction with AI helped someone engage more deeply in the material and created a stronger work product at the end of it. So I think these are all solvable problems. we are engaged in these conversations every single day and we want to be part of the solution

17:39still ahead i talk with rahul about how he thinks ai will shape the future of work

17:56real leaders don't back down when the stakes are high they innovate they push forward and then they take the stage at the masters of scale summit join us in san francisco october 7th to 9th to hear from the CEO of the New York Times, scientists using cutting-edge technology to find cures, the leader of crypto powerhouse Coinbase, a retired four-star general, and many, many more. Apply now at mastersofscale.com slash apply25. That's mastersofscale.com slash apply25. If you're ready to take your startup from idea to impact, then AWS is your launchpad. From data storage to machine learning to secure app hosting, AWS gives you the tech trusted by the world's fastest scaling startups.

18:46And here's the best part. Join AWS Activate today and you could score up to$100 ,000 in AWS credits tailored to your stage and network. Go to aws.amazon.com slash activate and start building. Expanding your business in the U.S. can feel like a maze. Every state has its own payroll, benefits, and compliance rules, which can pull your focus away from growth. That's why founders use Deal. Deal is the professional employer organization, or PEO, that gives you a dedicated HR expert plus Fortune 500 level benefits for your team. The National Association of PEOs says businesses can grow twice as fast if they use one.

19:30So if you're scaling, make it simple with Deal. Go to deel.com slash mos and get up to three months free. Welcome back to Masters of Scale. You can find this full interview and more on the Masters of Scale YouTube channel. You reference trust in our conversation, and I'm struck by the framework that you've developed that goes by the acronym TRUE. Trust, responsibility, user control, and empathy. Can you talk about how that framework came into use and how you actually employed at Grammarly? You know, it reflects our existing practices. It was just honestly a reflection of what we already did. It was a helpful way to frame all the things we were doing to evolve our product, but do it in a way that was true to our values.

20:29So for example, trust is critical. And I think when we created the true framework, it was pre-LLM scaling, but it's more true now than ever before. So all of our investments in a multi-layered security program, all of our investments in privacy and user control, all of those things are critically important now, in fact, more important now than ever before. So it helps us to have a North Star in terms of what the key capabilities are and the key values are we want to keep in place because it's a very fast-moving landscape and you need some framework to make decisions. So the true framework is our framework to make those decisions.

21:08So as an example, there's an open question with AI in terms of, well, is this data going to be used to train the model or how do I think about my data? and because trust is a key pillar of how we approach our work and because that's been true from day one, we have the opportunity for every single user of Grammarly, free or paid, individual or business, to opt out of using their data to train Grammarly's models. So that user control is part of what we mean by trust. So there's meat behind each of these things. It's not sort of an empty framework, but something we use in practice every single day to make decisions.

21:49And you alluded to this earlier. One of our co-founders, Jay Punjabi, says all the time how this is the worst the AI will ever be, right? Every day is literally the worst it will ever be. It's getting better by the day. And at the same time, it's not only not linear, but there are dips, right? So we've seen, for example, AI agents return results about how to keep cheese on pizza by using glue because that was in a Reddit post that made its way into the data set that the LM was trained on. So I'm curious, when you look at what's happening in the industry, there's a halo effect when everyone's employing a true model.

22:29There's almost a devil effect when they don't. And the media obviously loves to latch on to these stories. So no matter how much of a tech optimist you are, these stories really deeply undermine trust in AI. So I appreciate that Grammarly is doing this. As you, as an industry leader, look across the landscape, how do we make sure that more AI companies are employing these core principles so that we grow in the right way? I think we as an industry need to take our responsibility to deploy powerful AI systems very, very seriously. This is a very profound shift in how we do our work. The scale of the shift, like all big platform shifts, I think it's a bit overhyped in the short term and it's underhyped in the long term.

23:20And so we as purveyors of this technology need to really understand what this means, how people will do their work, what it means to knowledge work in general, and the impact it has on knowledge workers. And we've got to take that responsibility very seriously. What's unhelpful in these conversations is very abstract arguments about existential risks posed by AI. These are tools that we can use to augment our potential. That should be the conversation. it's actually quite prosaic. The set of issues that you need to think about are quite prosaic. How good is the quality of the training data? What specific investments are you making to make sure the training data is high quality, it's been annotated well, it's been quality checked?

24:03How are you evaluating the output of the model? How are you building classifiers to understand sensitive text and take corrective action? We have built, for example, a sensitive text classifier. It's a project that we call Seismograph, and we've open-sourced it. So imagine you're getting a bunch of text, maybe someone has written the text or maybe it's output from an LLM and the text may be sensitive. Maybe it's got hate speech or it's got some other sort of issues. And so we built this classifier called Seismograph that does that sensitive text detection. And so as an application, you can look at that output and say, it looks like this text is sensitive.

24:42I can now take some corrective action. I can suppress that so the user never sees it or I can go back to the LLM and say, that text wasn't great, give me another version. So we built this, we found it very useful for our users to keep them safe. And because we want to contribute back to the open source community, we open source Seismographs to anyone who is facing an issue of sensitive text. So we hope that other people in the community can take the work we have done and deploy it in their own AI applications. And so I think we can share best practices because as an industry, we must make sure that this technology is deployed in a way that augments human potential.

25:22How much is open source part of your next phase of scaling in your vision of this? Open source is going to be critical. So first of all, as Grammarly, today we're doing a bit of open source, a bit of closed source. We have a set of different models we're using, but long-term, I would like to bet on open source as a future of AI. Now, setting aside Grammarly, I think just as an industry, given the power of the impact that AI can have on how we do our work, I think the sunlight that open source brings is something that we as a society, as an industry, absolutely should be pushing for. Where does your belief in open source come from?

26:03How did that originate? I've just seen it. If you look at Windows and Linux, all of these data centers that are powering compute, they're all based on Linux. if Windows had won the Windows versus Linux battle, the world would have looked very different and I think would have been less able to scale and take advantage of this new AI desire for compute. I see this with open source web products. This is my firsthand knowledge. But in general, I think there's a lot of fear that gets created because transparency is scary. But ultimately, that provides us the best way to harden our systems. Because every tool can be used for good or evil.

26:48Tools themselves are amoral. Transparency is the surest way that I've seen in the history of tech to really harden those tools. Yeah, and yet you can also, I'm sure, appreciate that particularly for people who are not working in this day in, day out, headlines are scary. And there's a lot of science fiction, which was seemingly decades off, which is now, it's manifesting in reality today. And one of those concerns is job displacement. I know you all, unfortunately, went through a reduction in force recently, had a round of layoffs as you prepare for this next era. How do you think about balancing your responsibilities to your team with building the best run, smartest company possible and making sure that AI is a net job creator and we don't fulfill some of these more dystopian visions of what AI can do?

27:43Yeah, I think, first of all, I do think that AI will be a net job creator. You know, back in the day, when ATMs were first getting widely deployed in the US, there was a lot of fear that bank teller jobs would be extinct, thing of the past. But the reality is what happened was ATMs allowed bank branches to be created at a much lower cost. Therefore, there many more bank branches, and therefore the need for tellers only increased. And tellers could then focus on higher value-added tasks instead of routine transactions. I think this is a pattern that we see over and over again with new technologies.