The next generation of digital assistant will be digital twin. You know sometimes you might be a very important meeting and I join by myself. For some other meetings, I might send my digital twin to join a meeting. My digital twin not only listening to the meeting over there but also can help me make the decision as well. That's the world where we're going.
Welcome to the Logan Bartlett show. On this episode what you're gonna hear is a conversation I have with co-founder and CEO of Zoom, Eric Yuan. Now, Eric and I talk about a number of different things related to Zoom, including the 45 different product releases that Zoom just announced this week, as well as what Eric sees as the future for Zoom in the AI world that we're moving into over the course of the next few years. We also talk about the features that Eric thinks any person that uses Zoom should be using. We also talk about how Eric prioritizes launching different features. And finally, I get a question answered that I've long since wondered about the insight that Eric had in the early days of Zoom that helped lead it into the business that it is today. Really fun conversation that I've long wanted to have with Eric that you'll hear now.
Awesome. Thank you, Eric. Great, great to connect. So, I guess I want to start off. I saw the keynote today for Enterprise Connect was entirely about AI. For people that maybe missed it, can you give just a brief overview of some of the stuff you announced?
Awesome. Yeah, so our chief product officer Sida, she was at Enterprise Connect. We announced a bunch of new innovations including around 45 features primarily, you know, around AI companion. You know, a year ago, we launched AI companion 1.0, which is more like based on generative capabilities. You know, we just announced the Zoom AI companion 2.0 with agentic skills, with an agentic framework and functionalities. You know, a lot of innovations around Zoom AI companion 2.0.
And what is, I guess, the shift AI companion shifting from or evolving from a personal assistant to being truly agentic? Can you give an example of like what that means?
Yeah. So a good example, you know, take a meeting for this meeting for example, right? So normally, you know, prior to the meeting, I need to prepare for agenda, right? You know, write down notes. After meeting, I look at a summary, manually create a task. That's the way it worked before. You know, with the Zoom AI companion agentic capabilities, essentially take this meeting for example, after this meeting is over, you know, AI companion will automatically create some tasks, you know, assign to you, send to you, and we'll track, you know, even after the meeting is over, right? That's, you know, an agent is doing something, you know, on behalf of us. Otherwise, we need to manually to create the tasks based on the summary, based on the discussions.
I guess there's been a host of different AI tools that have popped up around meeting transcription and meeting summarization. And how do you think about the role that you play as an ecosystem hub for players like that versus building your own solutions in different areas that maybe relate to Zoom?
Yeah, great question. First of all, you know, we are offering an open platform. You know, third party vendors, they can build transcription tools by themselves, right? You know, given we have API. At the same time, you know, some of our customers, they would like Zoom to offer a seamless experience and everything from Zoom. That's the reason why we also offer the similar capabilities. Again, this open platform, you know, up to customers to make a decision if they want to use Zoom built-in, you know, all those AI capabilities, you know, like real-time summary or meeting summary creating tasks after meeting is over. You know, we offer those features. Or customer, they can go to any third party, that's okay too.
Yeah, my understanding is you all have been pretty purposeful about not charging for a number of the AI features that you guys have rolled out, which I think stands in contrast to a competitor of yours, Microsoft, who I think does charge for a lot of different stuff. How do you balance the democratization of AI and the thought process around embedding some of this stuff into your core platform with maybe the potential to charge for these things because they are so valuable?
Yeah, great question. So before we launched AI companion, we already spent a lot of time on talking with the customers how to leverage AI to add more value for the customers. And, you know, AI companion offers lots of AI capabilities. Those features or capabilities are core, you know, to our meeting business, you know, and we look at everything from a customer perspective, you know, what's the total cost of ownership, right? So that's the reason why we, you know, we truly optimize our back end, you know, and AI infrastructure and try to do all we can to make sure, right, the cost is very manageable. With that, we can offer a great service at no additional cost to our customers. That's the reason why AI companion usage, you know, is growing very well, like quarter over quarter 68% growth. And because of at no additional cost, because of the greater functionalities. At the same time, you know, some of the enterprise customers, they would like to customize AI companion, right? And to leverage our AI Studio, we can monetize on that front. But the core AI companion functionalities, no matter how we innovate, we still want to, you know, give customers at no additional cost.
As you think about these AI enhancements across Zoom, launching 45 of them, were they all 45 were announced?
Yeah. Yeah. Not every 45 features are just AI companion, and, you know, primarily AI companion related, right? For sure, we launched some other vertical services like Zoom Workplace for clinicians, for frontline workers, and so on and so forth.
Is there one to two things as you started exploring AI and some of the use cases that were maybe counterintuitive to you? Like when you set out sort of thinking about these things and the potential for Zoom to play in the world of AI, a conclusion that you reached that was maybe not totally obvious from the onset.
Yeah, one feature, you know, we are a real-time collaboration communication company, right? Take a Zoom call for example, you know, everything is around real-time, you know, collaboration. And one feature, like, hey, what if, let's say, Logan, you know, we are sitting together in a local Starbucks coffee, not on a Zoom call, when we meet in person, how to leverage the AI? So we launched a feature called voice recorder. Essentially, you know, I just put my phone on the table and the one-click launcher starts recording, right? And we automatically record our conversation, generate transcription and the meeting summary, and assign tasks, and so on and so forth. That's a pretty useful feature, right? And even beyond a Zoom call. That's one. Another one, really, you know, it's my personal favorite feature, which is calendar management. You know, like you and I want to schedule a meeting next week. You know, the way it works today, I need to call my EA, right, and talk with your assistant, right, to schedule a call. You know, a lot of steps. What if we leverage AI companion? Hey, AI companion, I want to schedule a meeting with Logan next week. You know, it will automatically give us options and make a decision for us. That's a very cool feature.
I assume a lot of these things are coming from the customers in some way, but how many of them do you think that more of these ideas sort of were surfaced in a bottoms-up manner from individuals that sort of saw this in the field and maybe they're an IC who had some observation that this is a way that it could work? Or were a lot of these like top-down from you or the leadership team and sort of setting the vision about where you ultimately want to go in collaboration with some of your big customers?
So, great question, you know, we are a 7,500 full-time employee company. Top-down push does not work. I do not have enough bandwidth to do that. Almost all the innovations are coming from bottom-up. So, meaning, you know, our innovation model at Zoom is always sit down with customers, listen to their, you know, feedback or pain points very carefully, and come up with a solution, and then talking to customers, this solution work for you or not. Almost all the innovations coming from that kind of a bottom-up approach.
I know you guys are very deliberate about, like, listening to your customers and trying to hear what they're saying or maybe finding novel ways of addressing the pain points that they have. Are there any lessons you've learned about conducting these customer interviews in trying to hear maybe not necessarily what the pain point is that they have, but what the problem that you can uniquely solve for them is? Are there any techniques or styles that you've come up with in sort of conducting these case studies or these interviews?
Yeah, wonderful question. So, yeah, for sure we have a formula here. Otherwise, you know, given our customers coming from online customers, SMB customers, all the way to the Fortune 100 companies, right? So, every day we receive so many feedback or feature requests, right? If you do not have a formula, sometimes really hard, almost impossible to prioritize those features. So, our formula here is more like, hey, when we talk to customers, we start from the problem statement. You know, human nature and customers would say, hey, you should do this, you should add that feature, right? You should build this functionality. You more like give a solution. But we always, we told our engineers, product managers, solution engineers, when we talk to customers, always ask what's the problem, and make sure we're on the same page about the problem statement or pain point. And then we work together with customer, try to understand what happened, what's the root cause. And then once we understand the problem root cause, we work together to come up with a solution, and then validate with the customers. And then we trying to build the product. Otherwise, if you don't understand the problem root cause, quite often customer say you do this way, yes we are going to ask you this way, another customer told us you should do that way. It's really harder to manage. That's why here we follow the problem, root cause, and solution framework.
How do you think about the areas that you have a right to solve those problems versus things that maybe are outside of your adjacency, like the calendaring example you gave? It's a problem we all deal with in a meaningful way and Zoom is the start of a lot of my or at least one of the first things I do in the calendaring process. And so I could say it totally makes sense for you guys to go down that path. I could also say there's a lot of things that are already integrated into calendar or the email stack or whatever it is and maybe that's an adjacency that doesn't make as much sense for Zoom to solve. And so how do you think about like where your scope exists and what is inside the boundary versus outside the boundary when you hear pain points from customers?
Yeah, again, you know, we take a calendar for example. We look at everything from a customer perspective. Customer told us, hey, they are using Google Calendar. Yeah, we integrate with Google Calendar very well, right? We build a calendar client, right? And we can talk with the Google backend, or we build a calendar client can talk with Outlook as well, right? At the same time, some customers say, hey, what if you offer the calendar native service, not only client but also server as well? We also offer the native client service. And some customers say I want to stay with Google Calendar or Outlook Calendar. We also have a plugin and to integrate with other third-party systems as well. Again, we know every customer is different. We try to understand a customer pain point and come up with a solution. Customer say I don't want to change anything. I want to stay with the Google Calendar. It's okay, we have a plugin with that, right? And that's the reason why our solution is very open and try to address different customers' requirements. You know, it's really hard for us just to build one solution, I do this way, can satisfy every customer's requirements. That's not how it works. That's why, you know, quite often, sometimes we might say, oh, this is our core, this one we can leverage a third party. You know, again, we want to make sure either we build it ourselves or we partner with a third party. We always try to offer a solution for our customer.
As you guys made this shift to being not just a video conferencing platform or kind of a core application of which video conferencing was one of the elements to more of an AI-first business, are there, is there anything you did specifically in the early days of that shift that if someone else is going through something similarly or trying to think about how to internalize the opportunity of AI, is there anything that you feel like you guys did uniquely well or made maybe mistakes you made that you would impart to a founder thinking about this?
You know, we are beyond a video conferencing company, right? So a few years ago, you know, we already talked about how to transform our business from being a video conferencing app to be an AI-first work platform. So essentially, we look at how customers every day they work, right? You know, open up the email, the calendar, send a chat message, make a phone call, schedule a meeting, and collaborate on document. You know, a lot of the tools, right? And customer need to back and forth go to the different tools to get work done. You know, we think about how to leverage AI to give customers Zoom Workplace, the entire collaboration suite. And where, like this morning, I open up Zoom Workplace, I can leave live, leave live, within Zoom Workplace, can get all of my work done. I check my email, I send you a chat message, right? I schedule a call, right? I do not leave Zoom Workplace. And also at the same time, leverage AI. You know, like used to be I need to manually create a document, now I can tell AI companion, please create a summary based on my last conversation with the partner and prepare me well for the next meeting. My AI companion more like my assistant, right? We try to leverage AI to look at each service we are building, how to improve that experience. You know, essentially, you know, more like AI will proactively help me get most of work done. I just assist AI companion to do that.
I know we have a lot of people that live in Zoom on a daily basis and so I'm curious if there's something that you've been using from an internal dogfooding standpoint, and maybe that was a good example, but anything that you would say, hey, if you're not using this within Zoom's platform, you're just really getting a suboptimal experience, you would encourage anyone listening to adopt this tomorrow.
I think a lot of things, right? And you take a meeting for example, right? And you take AI companion, like if I'm late to the meeting, you know, I just turn on AI companion, catch me up, right? It's a lot of features already available there, right? You know, like live translation as well. A lot of very cool features already built-in, right? I want customers to explore all those cool features. At the same time, a lot of new services. Again, you know, take Zoom Docs for example, right? I think it's a very powerful tool. You know, it's not like a Google Docs, not like the Mac of the world, right? I can create the knowledge base wiki and created a table within Zoom Docs track what's going on. After meeting is over, we automatically create a Zoom Doc, right, to track the meeting summaries and also the tasks and so on and so forth. You know, another feature I really like which is called Zoom Clips. And very soon we are going to have a feature that's also we announced at Enterprise Connect where, you know, I created so many videos, right? I personally created videos. You know, it's really costly, you know, how to leverage the AI. I just create my avatar and 2-minute avatar, I give to our marketing team, and down the road they just use a script automatically generate a video for our next earnings call. I'm going to do that. With that, I do not need to pre-record anything. I just send the marketing team the earnings call script, you know, we automatically create a video. So a lot of features like that is very cool and well beyond a video conference platform. And unfortunately, you know, take some time, many of our customers they did not realize. Another example, we have a built-in chat functionality, we call it Zoom Team Chat. You know, something like Slack or Teams. We had that for many, many years. You know, a lot of bigger customers also deploy that. It's a great building service. And, you know, we created a, you know, here I live on Zoom Team Chat, right? I send all the chat messages with my team, with all hands, you know, after all hands meeting. And we also have every meeting, after meeting is over, we create a persistent chat group. I can continue engaging with our employees. There's another powerful tool, you know, a lot of our customers they are not aware of that.
Customer education is obviously, as your surface area increases, it obviously becomes more and more of a challenge for the sales people, for the product team, for product marketing, whoever it is. Have you found that doing one big release or narrowing the number of times at which you're releasing and announcing features has led to better customer education? Or do you guys still ship when the product's ready and work on catching up and communicating later on to the customers as needed?
Yes, great question. So given the size of a company, given so many customers and the speed of innovation is still extremely important. You know, we just, that's the reason why every time I have so many, you know, new features. Because again, we are a workplace platform and also we have services like contact center, you know, a lot of new services and new features. The key is really not about those innovations. The key is how to, you know, easily help customers discover those features and make sure at the right context, right? Like take Zoom Docs for example, you know, customer, they are not aware of this feature. What if we think about a use case where after the meeting is over, we automatically created a Zoom Doc where you have a meeting summary. With that, we can expose those new services, new features. So we got to think about from customer's perspective how to embrace those new features rather than we leverage marketing, you know, that may not work given the so complicated services we build out. Because we have so many new things, right? So that's kind of the effort we are working on. And to, you know, take a meeting for example, right? After the meeting is over, we should tell you where to find those meeting summary templates rather than before the meeting we should tell you, oh, you should use a meeting summary template, how to customize, so on and so forth, right? So that's kind of our strategy.
As you zoom out and sort of look forward, what do you think the future of meetings looks like? Maybe in, I don't know, 3 years, 5 years, 10 years, whatever time horizon furthest out you feel comfortable projecting. Are there any things that you've internalized that maybe wouldn't be obvious to just the average person going about their day using Zoom?
I think the future, just look at a meeting experience. I am very excited about the potential and innovations, you know, like a few years later, right? So, you know, how to make sure every meeting experience very immersive. You know, more like you and I sitting together in a local Starbucks coffee, right? And very immersive 3D experience. And also, you know, if I shake your hands, if you can feel my hand shaking, that's even better. That's truly immersive. That's one. Another thing is given the AI evolution, right? We are building the digital assistant, but the next generation of digital assistant will be digital twin. You know, sometimes, you know, might be a very important meeting and I join by myself. For some other meetings, I might send my digital twin to join the meeting. My digital twin not only listening to the meeting over there but also can help me make the decision as well, right? That's the world where we're going to lead, right? You know, you have AI, you have AR, and together we will make the meeting experience very immersive, very intelligent.
So at a literal level, so that people understand it, I think unpacking the concept of a digital twin would be helpful. And so can you just go through at a very tactical level, like what you mean by that? Is it going to be avatar-based? Is it the decisioning side? What is it trained on? Can you just give like a little bit more granular insight into it?
Sure. I think, you know, more like a two-step approach. The step one is more like a digital assistant based on my digital avatar and also all the information I can access. You know, my digital assistant can access as well. Essentially, I have a personalized RAG, and also my email, chat message, meeting transcription, any information I can access, the data also be fed into my personalized area. You know, it will be part of my digital assistant. That's more like step one. The step two, another, a lot of knowledge here, you know, how to make sure, let's say, I download all my knowledge, my brain, right? Somehow to figure out a way to train my digital assistant, then my digital assistant would be evolved into a digital twin, right? I think that step two, I do not think ready yet. We might need some neuroscience breakthrough. I think that's kind of a two-step approach.
And how do you think about resourcing against something like that that might feel far off to the average person and the digital twin? Like how do you think about the allocation of resources or R&D development to something like that versus the standard product that you're selling today?
Again, you know, most of resources, you know, are working so hard on addressing customers today's needs, right? And maybe the next few months, few quarters. For sure, we have some engineers, you know, work on tasks for the future, right? Maybe, you know, may not be ready for one year, two years. You can call that a lab, right? So, you know, more like cutting-edge technology, right? And it targeted for the future. So, more like, you know, and most of the engineers for sure, they build foundational technology, you know, and more features, you know, AI companion, so on and so forth. We also need to think about the future as well. I think probably 90%, you know, are working on today's features and services, you know, 10% reserved for the future research work.
You guys obviously have access to a bunch of sensitive information and trust is such an important point of what it is that you guys do from a security standpoint, data, all of that. There does seem to be this tension that exists with a lot of the AI products today of a desire for the copious amounts of data that people can access with the need to ensure user trust and confidence with either existing or potential customers. I guess can you just speak to this and how you guys think about where the bright lines are that exist between the data that you want to train on and enable versus the data that you keep totally separate and is totally sovereign to the individual customer.
Yes, great questions. First of all, Zoom is a very, very secure platform. I take security very, very seriously. That's the reason why today look at almost all, most of the security companies worldwide, you know, they all use Zoom because it's very secure. But, you know, take AI or data privacy for example, we are the first vendor who made a commitment, we do not want to use any of our customers' content to train our AI model, period, right? So we made that commitment. So, you know, we just, we have, we build our own large language model but use, you know, and we use public data or maybe partner with third party, we never want to use our customer data. You know, that's kind of very, very clear. And if customer, they want to, like, turn on this AI companion, they want to get the meeting transcription, we also very, very flexible, you know, the policy. Like a customer say I do not want to have any data after meeting is over, I just needed the meeting summary. We have a zero data retention policy, right? Automatically deleted all the transcription, so on and so forth. And some customer say I don't, I want to make sure this meeting extremely secure, I do not have any companion. It's okay, we support the end-to-end encryption, which is the first vendor, you know, who claimed the post-quantum end-to-end encryption. We also support that as well. Again, data privacy extremely important. And we look at everything from a customer perspective. We don't want to make a decision for customer like, hey, for this meeting transcript, we should keep for 3 months or 6 months. No, we give a customer choice. Let a customer make a decision by themselves.
We touched on it a little bit earlier about your obsession with the customer. I'm curious was there an example in Zoom's history maybe of a product that became important where customer feedback maybe directly altered a direction that you were headed from a product roadmap standpoint in an unexpected way.
Yeah, it's happening every day. You know, I give one example like last night at a dinner with a customer. And again, you know, our AI companion is based on a federated AI approach, meaning we have our own large language model with Llama or OpenAI and Anthropic, you know, and this is a federated AI approach. And some customer, especially for financial institutions, they say they want to just use Zoom AI companion, you know, our own model. We give them a choice, or you can use all the models together as hybrid. There's some customer, I just want to use Zoom AI companion, maybe with Anthropic, I just use those two. Do you support that? Today, we do not support that. Either Zoom own large language model or federated approach. When we get those feedback, customer say, you know, they want to pick Zoom own large language model as one of the club-based large language models at this moment, do not support that. But based on feedback, we are going to support it very, very soon. You know, those kind of feedback we receive every day. That's the reason why our innovation speed, you know, truly help us stand out compared to any other competitors. Based on customer feedback, we can quickly innovate and very soon we're going to give a customer flexibility. They can pick whatever large language model they want to use.
In terms of your stylistic approach to running the business, there's been this refrain around, I think founder mode was the term of Paul Graham and I think he attributed a lot of the distillation coming from Brian Chesky. And for, for lack of a, for a good synthesis of at least how I think about it, it's empowering the founders to sort of make centralized and maybe gut-based or forward-looking decisions that ripple through an organization, which I think stands in contrast to a lot of executive empowerment and probably elements of delegation as well. I'm curious, do you gravitate to being more hands-on and in the weeds of the individual functional areas, or do you tend to stylistically approach it more from a delegation standpoint?
Yeah, I like that conversation between Paul and Brian. It's really, you know, love that. I think most of founders, I think probably, yeah, in my view, is extremely hands-on. And I know from the day when I started, you know, 2011 until today, you know, still very, very hands-on. You know, and sometimes you need to delegate, quite often, you know, it's hard to delegate, right? You really, as a founder, right, you know, let's say, you know, when customer ask us a problem, you know, I really look at it from a customer perspective. I could have delegated to my direct report and then delegate to another manager and then talk to engineers. I cannot wait. I immediately assume a call with one of engineers, try to, what's the problem? When can we come up with a solution? Can we deploy this weekend? This happening every week, right? Like a founder, I, you know, is, you know, truly care about everything. And then also the speed is very important. That's the reason why sometimes you have to be extremely hands-on to talk with the individual contributors and get it done as quickly as possible. Because, you know, every day customer might be suffering from those issues. You know, that's the reason why you have to be extremely hands-on sometimes, a little bit micromanagement. But again, the goal is to make sure deliver happiness to our customers. As long as I have that goal, I really do not care about delegation or not a delegation. I just want to get work done. That's it.
You all have gone through different journeys over the course of the last couple years. I guess every business has, but you guys uniquely certainly had, were enabling...
Infrastructure for a lot of us just to survive with friends and family and then also do our jobs over the course of the pandemic. I shudder to think of what the economy maybe looked like absent Zoom for all the stuff that you guys were able to enable. And then there was sort of this zero interest rate phenomenon that we all went through and then we've sort of had the hangover of all of that stuff that's happened since then. I'm curious as the dynamics in the market have changed, how has your culture evolved? How have you thought about shifting the internal values or whatever the case? Or maybe not. Maybe you've kept them consistent all the way through. I'm curious like in going through these different cycles how you thought about what to iterate on versus what to keep consistent from a cultural standpoint.
Yeah, Logan, that's a wonderful question. You know, speak of a culture, that's one of the things that keep him up, you know, in the night. The reason why, you know, prior to pandemic crisis, you know, we were relatively small company, 2,500 employee company and I know almost all of them, right? So and we had a great culture and every day we're working hard to truly deliver happiness to customers, look at everything from a customer perspective. You know, whenever we got any escalation, everyone want to say, hey, can we deliver today? Can we deliver this week? We have that mentality.
But during COVID, you know, crisis, right, given the service demand, we had to recruit so many new employees. Essentially, within 18 months window, we hired around 6,000 employees. On the one hand, we had no choice but to hire many employees. On the other hand, we made a mistake. I made a mistake. We hired new employees too quickly. I can tell you the company culture was broken a few years ago. You know, we are almost, you know, fixed the company culture problem, right. The reason why is, you know, sometimes we do not look at everything from customer perspective, always from look at from internal perspective. Oh, we should add this process. Oh, we should do this, do that. We forgot everything we do here. We got to make sure customer feel happy and we lost that. And notice, you know, we are trying to fix that problem. You know, is, you know, recently we just published our culture playbook. The goal is to make sure everyone really understand why we're working so hard here, what's the goal. You know, our company value just one word which is care. You care about the community, customer, company, teammates as well as ourselves. If we do not care about the customers, no matter what we do, we think we have innovation, customer, they do not care. What they care is what we can deliver our services and more value to our customers. When customer use our service, they feel very happy. Wow, it just works, right. That's the reason why, you know, we are working so hard. And again, you know, we are almost there and to fix the culture, bring our, you know, previous culture back. We are not there yet. Hopefully by the end of this year. I told our team it might take three years for us to bring the greater culture back. This is the third year.
