Zoom CEO Eric Yuan wants AI clones in meetings

3 Jun 2024

I had the pleasure of speaking with Eric Yuan, the CEO of Zoom, and to my surprise, our conversation veered off in an unexpected direction. Eric launched Zoom after working for Cisco, where he observed an opening to streamline and simplify video conferencing processes. His foresight proved correct, as Zoom has now become a widely recognized name - particularly due to the surge in its usage during the COVID-19 pandemic.

However, the usage of Zoom has decreased recently and as a result, the company is now dealing with a variety of business issues that were discussed between Eric and myself. Nevertheless, Eric has a much bigger plan for Zoom beyond just being a video conferencing platform. He aims to compete with Microsoft and Google in the business software industry by adding productivity tools such as email, document editing, and chat. Like many other businesses, Zoom has also invested significantly in AI technology, and Eric has some ambitious ideas for what this AI can accomplish.

Tune in to Decoder, a program led by Nilay Patel from The Verge that talks about major concepts and other issues. Sign up for it now!

Eric strongly suggests that you stop subjecting yourself to Zoom meetings. He believes that artificial intelligence will bring numerous advantages to the workplace, such as the ability to create a "digital twin." This refers to an AI-generated version of yourself that can attend Zoom meetings in your place and make decisions on your behalf. You can now use your time more productively, for example, by spending time with your loved ones.

Let me give you a heads up: I was attempting to inquire about the typical Decoder queries throughout this discussion, however, I ended up asking plenty of additional questions when we started talking about digital twins being present in Zoom meetings for individuals. How many digital twins could there be? How would they manage to stay coordinated with each other? Are they reliable? What work is left if everyone dispatches their digital twins to every meeting?

Eric was very willing to discuss these concepts with me, leading to a CEO interview on Decoder that was quite distinct. I haven't been able to stop contemplating it since we conducted the interview. I believe you will find it enjoyable.

Let's talk about Eric Yuan, the creator and leader of Zoom.

The following conversation has been slightly modified to make it shorter and easier to understand.

Hello Eric Yuan, it's a pleasure to have you on Decoder as the creator and top boss of Zoom.

I want to express my gratitude for inviting me. I value and am thankful for the opportunity.

I am thrilled to have a conversation with you. It seems like the software industry is experiencing some interesting changes due to AI and you are making a significant investment in this area. During the pandemic, Zoom stood out as a major success, gaining widespread popularity and acceptance. Now, you are making substantial changes to the product, which is fascinating and worthy of discussion. There is so much to explore and discover.

To begin with, we'll start from scratch. Zoom is widely recognized as a videoconferencing application. A number of fresh features have recently been launched by them. This includes characteristics suitable for workplaces as well as Artificial Intelligence (AI) features. How do you consider the current state of Zoom?

I believe that we are currently beginning a new journey towards version 2.0. You are correct in your assessment. When we reflect on version 1.0, it was primarily focused on developing certain applications, with video conferencing being one of them. Our motto was "Work Happy." However, the focus has now shifted with version 2.0. It's now "Work Happy with the Zoom AI Companion," which encompasses the entire collaboration platform along with AI technology.

Recently, I've come across a similar tale from multiple enterprise software providers. The one that stood out to me the most was Squarespace. Despite being known as a web-based company, they claim to take charge of running your back office while the website is simply a way of scheduling. Why is it that this appears to be a common trend among enterprise companies? Is it because they strive to expand and dominate various aspects of office operations after developing one flagship feature?

You are absolutely correct. This situation is likely applicable to nearly all businesses operating as software-as-a-service providers. It is important to understand that these companies are all offering a form of service. With the advent of artificial intelligence, however, there is a growing focus on how to better utilize AI in order to automate processes, minimize the need for manual labor, and decrease the necessity for human intervention.

When considering the brief introduction for Zoom, you reflect on when the creator had to gather funds for the platform. Initially, it was straightforward. Videoconferencing posed significant challenges due to the specialized equipment and cost. However, Zoom was designed to function as a user-friendly app, making videoconferencing accessible to everyone. So, what is the current concise explanation for Zoom?

Regarding conferencing, there is this one application that offers more than just the service of connecting to a video meeting. It also lets you read and write emails, chat with others, make phone calls, use a whiteboard, and organize events with people outside your organization. What we are currently doing is examining your complete schedule in order to make the most out of Zoom Workplace, which can assist you even when you are not using it. Our goal is to help you accomplish most of your work with Zoom Workplace. We still have some work to do to make this a reality.

For this session today, I don't necessarily have to physically attend. I can instead use a digital version of myself to attend while I enjoy the beach. I also don't need to worry about checking my emails as my digital self can handle most of them. Only a few may require me to personally respond. Nowadays, we spend a lot of our time communicating through phone calls, meetings, emails, and text messages. It can become overwhelming. That's why it's crucial to explore ways to automate these tasks using AI and tools like Zoom Workplace. By doing this, we can free up more time for other important activities.

Our blog is followed by a vast group of individuals including product managers, engineers, and designers. As per my understanding, they plan to use AI avatars for their daily stand-up meetings.

Additionally, it's not just for conducting meetings. It extends to my email correspondence as well. To be honest, I really despise going through my emails every morning and ideally, my AI doppelgänger would handle the brunt of it. Unfortunately, we haven't quite achieved that level of technological advancement yet.

The idea of automating various tasks in the workplace, such as status updates and non-decision related conversations, is a significant goal. This would allow for a version of oneself to communicate information quickly or to automate email processing. It's a common dream, and we'll discuss how we're working towards it with Zoom. However, if AI handles these tasks, what will be left for the average worker to do?

As it stands, AI is not yet capable of fully taking over tasks and it will likely be some time before it is ready to do so. However, looking ahead by five or six years, it may be able to assist with up to 90% of the work. Nevertheless, in terms of real-time interactions, technology is limited. For example, when we communicate online, even with digital versions of ourselves, it cannot fully replace the experience of meeting in person. When we speak face-to-face, there are physical actions like hugs or handshakes that cannot be replicated by AI. Therefore, there will always be a need for personal interactions, such as meeting in a café and having a meaningful conversation. These are significant moments that cannot be replaced by technology.

The central component of Zoom remains its videoconferencing service. This is how I perceive it, and it's how the majority of people perceive it. Are you stating that Zoom's videoconferencing tool will primarily comprise AI avatars, and will it encourage us to hold more face-to-face gatherings?

I have two perspectives to share. Firstly, our company is not limited to just videoconferencing. We offer many other features that make up the entire Workplace platform, which serves as a collaboration tool. Secondly, when it comes to videoconferencing specifically, there is potential to utilize more AI technology. This would help save time wasted in lengthy meetings and reduce the need for multiple Zoom calls throughout the day.

We could have the opportunity to spend more time interacting with each other face to face, but not necessarily for work. It could be for other activities. Why is it necessary to work for five days straight? In the future, we could consider reducing it to four or three days. Then we would have more opportunities to be with our family, focus on our creative pursuits, and give back to the community and society by helping others. However, at present, it is difficult to achieve this desirable goal because we are tied up with our busy work schedule every day, for five days every week. This monotony can be dull and tiring.

To clarify, this discussion is fantastic. If I understand correctly, you believe that Zoom will help us achieve a four-day workweek by utilizing artificial intelligence for our video conferences?

Definitely! We rely on various methods of communication for our daily work, such as Zoom meetings, chatting and messaging, phone calls, emails, utilizing whiteboards, carrying out coding tasks, handling creative work, managing tasks as a manager, and maintaining project management. To achieve efficiency and improve performance, we incorporate the use of AI-powered tools and innovative applications. This is the path we are taking as part of our Workplace platform, which represents our journey to version 2.0.

I've got a ton of inquiries, but let me pose this one initially. When considering your personal schedule, envision yourself as the chief executive officer. Undoubtedly, you have a plethora of tasks to complete, making you the most occupied person. Assuming that AI had the capacity, what duties would you delegate to it today?

I have numerous thoughts about my day-to-day schedule. Initially, I must express my strong disliking towards my calendar. Every day, when I glance at it, I feel overwhelmed by the number of tasks looming ahead. It's already evident that I will have an overwhelming amount of 9 to 10 meetings. In the midst of it all, I must make time to read messages, check emails, and take phone calls. This makes me feel uneasy whenever I look at my itinerary.

The next thing to consider is how I ended up in this situation. Typically, it's due to my administrative assistant scheduling various meetings, and occasionally I'll schedule meetings myself. It's important to note that scheduling a meeting is not a quick and easy task. It requires coordinating many different aspects. The third factor to consider is analyzing all the meetings on the calendar and determining which ones are mandatory and which ones are not. Additionally, it's crucial to avoid overbooking oneself.

Can you tell me which meetings you attend and consider delegating?

To begin with, I addressed the issue at hand. However, what happens if I am unable to attend a crucial meeting due to my busy schedule? I may struggle to comprehend what occurred during the meeting. Furthermore, being the CEO, if I miss an important meeting, it may have to be postponed as I am the decision-maker. Without my presence, progress cannot be made, and the meeting would need to be rescheduled. When we take into account all of these challenges, we can imagine the benefits of having AI support. If AI had access to my calendar and understood the context, it could effortlessly schedule a meeting with just one click, enhancing efficiency significantly.

Every morning when I rise, an AI informs me, "Eric, you have a total of five meetings today. It's unnecessary for you to attend four of them. You need only attend one. You can transmit a digital version of yourself." For the sole appointment I do attend, I can obtain a condensed report and disseminate it to those who weren't able to attend. It enables me to make better choices. Once more, I can utilize the AI as my aide, and it can give me a multitude of advice beyond my capacity. This is the objective.

I am assuming that when you checked your schedule today and noticed a Decoder session, you had planned to attend it by yourself. However, hypothetically speaking, if you couldn't make it to the session, would you have sent an artificially intelligent (AI) avatar in your place?

In my view, an AI avatar is essentially an artificially intelligent version of myself, isn't it?

Basically, to be able to listen to a call and engage with a participant in a significant manner, my digital twin can represent me and participate in the decision-making process. However, this is not yet possible due to the limitations of the current LLMs. Currently, everyone uses the same LLM, which doesn't make sense. Each person should have their own LLM, like Eric's LLM or Nilay's LLM. This is the foundation of the digital twin, and it allows me to rely on my digital twin to participate in meetings or conversations when necessary. In the future, I'll be able to choose whether to join a conversation myself or send my digital twin to participate on my behalf.

What do you reckon is the timeline for us to achieve that future?

In my opinion, it will take a few more years for us to achieve our goal because there are two main issues we still need to address. Firstly, the current large language model is relatively new, and although it holds a lot of potential opportunities, it hasn't reached its full potential yet. Secondly, it's important that each individual has their own personalized version of the LLM based on their unique experiences and context. This means that every person will have their own custom LLM, and some may have multiple variations based on their strengths and weaknesses. For example, if I know I struggle with negotiations or sales calls, I can tweak the parameters of my LLM to better suit those situations. Overall, we're just getting started on this journey and there's still a ways to go before we reach our destination.

Do you ever wonder if your dial would read something like "improve sales skills"?

Yep, for that gathering, I recommend saying, "Hey, adjust that parameter to improve your ability to negotiate, send over that updated version, and then join in."

If you contemplate this idea as presented on Zoom, the online video chatting application, do you believe that an animated version of yourself in 3D, such as the Vision Pro face features created by Apple, will be involved, or do you assume that only your voice will be included?

To begin with, the primary mode of interaction might be voice recognition. However, in the future, the technology will evolve to become even more immersive. We can already see this happening with advanced devices like Vision Pro and Meta Quest 3. This is just the start of something big, and in the days to come, we can expect to see a 3D representation of ourselves that will be so lifelike it will be difficult to differentiate between a real person and the 3D model.

Wow, there are heaps of complicated technical issues to tackle here, aren't there? We've got this cool 3D avatar that looks ultra-realistic, and then there's this LLM thing that might work better if we adjust some of its settings. But honestly, lots of folks still don't trust it because it messes up sometimes. Plus, we're trying to make sure that our digital avatar feels culturally relevant to people all around the globe. It's quite a daunting to-do list! So how is the Zoom team going about solving all of these challenges and bringing our dream to life?

In the past, we discussed the goals for Zoomtopia, our conference for users. We envisioned a world where distance doesn't matter. For example, if I live in San Jose and you live in San Francisco, we could still have a virtual meeting that feels like we're in the same room. We could even have the ability to shake hands and hug each other through technology. Even if we speak different languages, we could use real-time translation to communicate without any issues. And if we can't physically be there, we could send digital versions of ourselves to have the same conversation. This was the vision we presented a few years ago.

So, how can we achieve that? In my opinion, we're already on the right track, which is great news. The industry is taking some steps forward. There are two technological advancements that are going to accelerate our progress. The first is Artificial Intelligence (AI), and the second is Augmented Reality (AR). Although we're just scratching the surface with these technologies, they have a promising future. Just take a look at Meta Quest 3 and Vision Pro. We're seeing some exciting advancements already. Even though both AI and AR still need some improvements, they will be instrumental in helping us reach our goals.

Lately, Zoom, as an organization, has undergone certain alterations. They had to let go of approximately 150 employees, which constitutes about 2% of their workforce. Due to this shift, can you explain how the company is currently set up to support its objectives?

As a company that is publicly traded, we prioritize discipline. Simply investing in more research and development won't cut it. We constantly analyze all positions and figure out ways to make the company as efficient as possible. This allows us to redirect some of the budget towards new and innovative technologies such as artificial intelligence. Many companies nowadays are seeking ways to save money as they invest in more graphics processing units, hire AI engineers, and recruit experienced designers and engineers to keep up with the latest developments in AI and understand the Vision Pro and Quest 3 world.

How can you ensure that you grasp the latest technology and put it to use in your current products? The answer lies in acquiring fresh skills. It's important to make your company more productive - after all, how else will you find the funds to support such efforts? This is exactly why we evaluate our positions every quarter to see if we can allocate some of our budget towards such innovative technologies as AI.

Therefore, this is what is causing the transformation: "We will cease carrying out certain activities. To our regret, we will have to terminate the employment of some individuals. We will channel resources towards H-100s and AI engineers." So, where do you stand? What is the current setup of Zoom?

In terms of structure, it's similar to the past but more effective. We began investing in GenAI even before ChatGPT emerged last year, which was done by many other companies as well. However, with ChatGPT's arrival, we recognized the need to increase our investment in GenAI. At present, we have significantly more GPUs compared to two years ago, and our team of AI engineers is also larger. Every day, our focus is on AI technology, the foundational AI technology, and our AI platform team. In fact, we have a lot more engineers than we did two years ago.

Do you have any teams apart from your dedicated AI team?

Our company has a team of experts in artificial intelligence, led by our Chief Technical Officer, who came from other businesses. Our company has multiple teams, such as Research and Development, who focus on introducing new services through gradual innovation. Additionally, we have a team dedicated to sales, marketing, and information technology. This growth in teams is necessary as many companies explore the potential of AI to enhance their products and streamline their business operations.

How can we ensure that AI is utilized in marketing material sales teams? AI can be utilized by sales managers to monitor and assist sales representatives during customer interactions after virtual meetings. Using AI, sales managers can identify productive sales sessions and gather insightful data. It is important for the entire team to adopt AI technology.

What's your strategy to introduce it? Are you announcing it to everyone like, "Guys, we need to do this." Or are you highlighting how the market is requesting it? Who or what is pushing the adoption of AI in your product?

In my opinion, there are two important things to consider. Firstly, it is necessary for us to keep a close eye on the advancements in technology. When we experimented with ChatGPT in the early part of last year, I was impressed and thought that it had huge potential, much like how people felt when the internet was first developed in 1995. It's important that we accept and embrace these technological advancements. Secondly, I devote a significant amount of time each day to speaking with customers and potential clients. It's interesting to note that the most common question I get asked now is, "What is your AI strategy and how do you incorporate it into your services? Are there any other AI-powered features you are developing?" This has made me realize that we are currently in the AI era, and not just ourselves, but our customers also prioritize AI when it comes to choosing a service.

If a customer like me were to ask you "What's your AI strategy?" what would be your most impactful response that would lead to a sale? What approach do you take to convert my question into revenue?

We want to focus on AI as our main priority for our company. We're looking at what we've already implemented, such as meeting summaries, and asking for feedback from our team. We're also integrating AI into our Contact Center and other services. Customers who have tried our AI Companion features are already enjoying the benefits. Personally, I've been using the meeting summary feature and it's been impressively accurate.

I am grateful for the mention of accuracy because it reminds me of the ChatGPT craze. When computers could communicate with us, we were amazed and thought it was incredible. However, now that many features have been released by different companies like OpenAI and Google, we have come to realize that not all of them work as well as they should. We are now on the other side of the excitement and are carefully assessing the accuracy of all the features.

The use of LLMs as a fundamental technology is convincing, however, they tend to produce hallucinations often, leading me to question their reliability. I have my reservations about whether the foundation to build the proposed vision is stable enough using this underlying technology. The challenge today with the grand AI visions is how to overcome this uncertainty. Although LLMs allow for expanded imagination, it may not be the most suitable technology to build a solid foundation for the vision.

I agree with you. It's impossible to develop a new technology overnight without facing challenges. One of the well-known challenges is hallucination, but with time, progress will be made to fix this issue. However, if we reject technology because of this problem, we may miss out on its benefits. Personally, I have not experienced hallucination with the meeting summaries, which are incredibly accurate. Although, it's possible the problem may arise in different areas.

The meeting summary works similar to an assistant for LLMs. You input data into it and ideally, your LLM can understand it from multiple angles, such as audio-only. This feature can help create a summary of the meeting. I comprehend how it could be beneficial in solving that issue. The ultimate goal is to have a digital twin at the meeting, making decisions for you. However, we have not yet found a solution for hallucinations, which is a crucial obstacle in achieving this goal. Even though model capabilities are increasing rapidly, it's unlikely that we will have a Moore's law-type solution for hallucinations, where we can anticipate reaching an acceptable level.

You're absolutely correct. That's why today, I'm unable to share a digital version of this call with myself. However, I believe that it's only a matter of time before this becomes a reality. It may require changes in architecture or the development of a new algorithm. It's reminiscent of the early days of the internet where there were many limitations. My phone was slow, and it hardly worked, but now we can see how far technology has progressed. Similarly, I believe that we will eventually overcome the current limitations, and these problems will be rectified. For instance, we already have powerful tools like GenAI, but we still don't fully understand how the human brain works. Think of what we could achieve if we were to unlock such knowledge and utilize it in GenAI. This is what the future holds.

like most people—don't think about the technical details behind the development of AI and how it will impact society. However, when you speak about solving the problem of hallucinations, it reminds me of the concept of transistor density in Moore's law. Other CEOs have discussed AI with me in a similar way, but I'm not sure if you have a deep understanding of this subject. Do you assume that Intel, Nvidia, TSMC, and other companies will continue to increase transistor density on chips, and thus make them more powerful for building more applications? Please correct me if I am mistaken, as I realize this may not be an area of interest for everyone.

I completely agree with you. Indeed, it's a technology stack that relies on numerous other components.

Is the issue of the AI model experiencing hallucinations located lower in the system, or are you actively working to reduce the frequency of such incidents?

In my opinion, finding a solution to the AI hallucination issue will be resolved in due time.

I suppose my inquiry is regarding the individual or entity responsible. Is it your responsibility, or does it lie further down in the hierarchy?

It's somebody lower in the hierarchy.

In my opinion, we need to focus on either the chip level or the LLM itself. However, our current focus is on the application level. We aim to improve the user experience by incorporating AI technology, creating innovating features, and supporting personalized LLM. This is our primary objective.

You stated before that everyone is utilizing the identical LLM. Is this correct? There are only a few LLMs that have the capability to perform the tasks you spoke about. Many applications simply utilize the ChatGPT API, or they may have recently switched to Gemini. Do you believe that significant distinction is necessary among these LLMs, to the point where you need your own model at Zoom?

Right now, when we look at all the LLMs available, it may seem like there are many options, but ultimately they are all pretty similar with only minor differences. We are just at the beginning of this technology, and I don't believe this is what the future holds. The future of LLMs is all about personalization. Each person will have their own unique LLM with different features and capabilities. Even every business will have their own specific LLM. It doesn't really make sense for all of us to have the same LLM because the reason why my LLM is so useful to me is that it understands me specifically. I believe that in the future, your LLM and mine will be very different from each other.

You began Zoom by recognizing a serious issue amongst customers - videoconferencing was a challenging task. Determined to find a solution, you developed an app to tackle this problem which has now evolved into a successful company. Your aspirations for the business are grand, and your insight has broadened significantly. You now speak of a significant shift in the world of computing and a cultural shift that will embrace the use of artificial intelligence. This is a far cry from simply addressing the difficulty of making a video call. In light of this metamorphosis, how has your decision-making process changed? What considerations do you now take into account when making crucial decisions?

It is completely true. You are absolutely correct. But we must also acknowledge the potential for growth and progress in front of us. To do this, we need to adopt a positive mindset towards change. We must ensure that our culture embraces the opportunities lying ahead. It is crucial that we continue to innovate at a rapid pace. Additionally, we need to collaborate more closely with our customers and prioritize our tasks effectively. Simply giving orders without consideration is not a viable option.

If we don't take heed of our customers, even if we have impressive technology and incorporate it, it's probable that our delivery won't be well-received. That's the second point. Furthermore, we ought to have an ambitious plan in place, which means committing resources for the future. Some of the things we're developing may not get released in the upcoming year or so, but it's critical to invest for the long haul. I believe these factors will enable us to excel and become an AI frontrunner.

Discussing investments for the future without seeing immediate results can be challenging, especially now that Zoom has become a public company. This means that you will experience pressure to report your progress every three months. How do you protect this part of your business from those quarterly expectations that come with becoming a publicly traded company?

Thankfully, we are making a profit. We are bringing in more money than we are spending, and we are following strict guidelines to ensure our success. You are correct that being a public company means we have to make adjustments to our spending. We need to consider what Wall Street experts think, and examine our profit percentage. We still have room for improvement in this area. However, we can redirect more of our resources towards projects that involve artificial intelligence or related technology. It won't be easy, but we can manage it. We need to accept these changes and move forward.

You are making more money than what you are spending - that means you are making profits. However, the pandemic is not as severe as it was before, and people are going outside. You are emphasizing the importance of physical interaction, but your rivals have implemented similar features as Zoom. For instance, Google Meet and Microsoft Teams are available and included with their office software. This makes it challenging to compete with them. What do you think? Will the videoconferencing industry prosper or decline?

When we observe the big picture, the industry of videoconferencing is progressing. The number of individuals utilizing videoconferencing is on the rise. Although you mentioned the opponents, interestingly, if you speak with individuals who have used a Zoom product, try the Zoom encounter, or experiment with our competitors' goods, you'll realize that we are still significantly superior. This is an important point to consider.

The second point to note is that there are many aspects that are still incomplete. We have the potential to be significantly innovative. Our latest release, Zoom 6.0, has numerous small yet noteworthy advancements which our customers enjoy such as multi-speaker view and animated interaction. Nonetheless, there is still a lot of room for enhancement. We can explore various ways to further improve the meeting summary by creating actionable insights for upcoming meetings. There are plenty of opportunities to escalate our innovation and we are far from reaching the limit.

Before starting Zoom, you worked in Cisco. Sometimes, you may have to use Webex. Have you ever felt furious while using it? Because that is how I usually feel about it.

I reckon that in the last year or so, I haven't participated in any Webex meetings. However, around a year back, I did attend one after receiving a link from someone. It's worth noting that while I did join on that occasion, I haven't done so lately.

No need to be concerned, there have been no updates or enhancements in the past year. What was your impression of the Webex meeting when you joined?

I sense a sense of contentment, like "Oh, I already own an iPhone, no? I am an iPhone user. You can offer me an old version like iPhone 1.0 or give me a Blackberry, but I have no interest in using them now."

That's great news! It seems that videoconferencing is becoming increasingly popular, with a growing number of people using it. The pandemic has definitely played a role in this, as more and more people have become comfortable with the technology. In fact, there has been a shift in cultural norms around videoconferencing in the workplace, with people using it for a wider range of purposes. However, the major players in the industry have really embraced videoconferencing as part of their overall bundles. If you want to convince an IT person or a CIO to opt for Zoom instead of the bundled version of Google Meet or Microsoft Teams, it may be challenging due to the current industry trend of prioritizing efficient budgets.

That's a great question. I don't believe a bundle pricing strategy is fair. I hope they can compete with fairness, but that's not under my control. However, we always consider the end user's perspective. I divide these customers into two groups: the first who prioritize cost over the employee experience or the quality of video service. I deploy their solution regardless of whether their staff likes it or not, just as long as it's within their budget. The second group is comprised of companies from Silicon Valley where employee satisfaction takes top priority. These companies always strive to deploy the best service that caters to their employees' preferences. For them, happy employees mean satisfied customers. We prioritize those companies and that's fine too.

Some companies don't just focus on the price of a product or service. Instead, they take a closer look at the total cost of ownership. This means that they examine all the costs associated with using a product or service over its lifespan. Zoom is a great example of this approach. The company has lower support costs than its competitors and offers an AI Companion to paid customers at no extra charge. By contrast, some competitors charge $30 per user per month for similar services. Considering these factors, it's clear that the total cost of ownership for Zoom is much better than its competitors. If you're looking for a reliable video conferencing solution that won't break the bank, Zoom is definitely worth considering.

Have you heard about Microsoft's bundle pricing issues in Europe? It looks like they'll have to unbundle Teams due to antitrust action. Are you keeping tabs on this situation? Do you support it or are you simply observing from a distance?

I take notice of this because we aspire to compete with integrity. At the end of the day, we all aim to collaborate with other businesses, even those we're competing against, to create the finest product for our customers. And who gains from this? Our customers do. If there's unjust competition, it has a harmful effect on the industry as a whole, and the final consumer suffers as well. For example, in the past, we all utilized Netscape; however, when Internet Explorer was introduced, it was arguably one of the most substandard software offerings in history and did not benefit users.

Do you reckon that separating Teams by Microsoft in the European Union will bring a beneficial effect on Zoom?

In my opinion, that seems to be the correct path forward, but there is still room for improvement. It appears that they will only be implementing these changes for new customers, which isn't fair to those who are already using the service. Additionally, even with the option to purchase individual features separately, the pricing still seems unjust. I hope they manage to rectify this situation. Unfortunately, I don't have any influence over their decisions, but it's important to realize that this is not a positive development for the industry as a whole.

On the other hand, there are some features of Zoom Workplace that already exist in other software packages and are being packaged together with Zoom. To explain better, let me offer an example. There is now a chat function that looks very much like what you would find in Teams or Slack. Additionally, Zoom now offers some meeting scheduling features and AI summaries. Zoom also appears to be moving into the realm of email, which is quite intriguing. Given that they feel that the existing bundle is inequitable, how do they plan to expand their current bundle?

That's an excellent inquiry! We implemented team chat since the very beginning, but unfortunately, we didn't promote it too well. It's quite sad since it's not like we add it later on, after some time. This feature was already available on the very first day. Additionally, let's take a look at our meeting scheduler. Admittedly, we do have some minor rivals in this aspect. Nonetheless, two compelling factors make us stand out: our clients can purchase the meeting scheduler separately, and our bundle cost is reasonable. We do not offer an extremely sparse price.

Take a close look at our range of services and you'll see that they stand apart from what our competitors offer. In some areas, such as email and calendar, there are only one or two dominant players and consumers have limited choice. However, our services present a different story with a wider range of options to choose from. While videoconferencing also has several providers, no one has a majority share of the market. In comparison to our competitors, we don't believe we engage in unfair bundling as our services are unique and varied in comparison.

I often speak with CEOs of enterprise software companies about a common concern: whether the customers, the CIOs, will opt for a complete package from tech giants like Microsoft or Google, or select a "best of breed" combination of their own. This ideal mix might include Slack, Zoom and various other applications jockeying for dominance. Slack’s huddles, for example, put it in competition with audio calls. Even if you go for the finest option available for each functionality, there is still tension among the products. How do you handle this challenge, knowing you fall in the best of breed category and aim to ascend to the big bundle level?

Zoom Workplace offers an open and flexible platform, unlike our competitors who often push their customers to use their entire solution. We don't believe in telling our clients, "You have to use everything from Zoom." Instead, we offer options that empower our clients to make their own choices. For instance, you can choose to deploy Zoom team chat or opt for Slack or Teams instead. You are also welcome to use our whiteboard or explore other similar options available. We value integration and have integrated with services like Zoom Clips and Loom, which gives our clients even more choices. Our platform is open, and we don't force our clients to use specific features they don't want or need. Unlike our competitors, we give our clients freedom of choice to create their own ideal workplace environment.

Expanding your business by incorporating AI elements may concern you, especially in terms of privacy and security. You are required to enter a significant amount of data in these tools, perhaps involving some form of training that is uncertain in nature and accuracy, followed by an AI assistant. Certain AI features on Zoom have already caused a stir due to the source of data used. Hence, what is your present stance on this matter?

When dealing with AI, it's important to act responsibly and be held accountable. This is why we proudly stand as the first AI company within the videoconferencing industry. Last year, we vowed to our customers that we wouldn't use their data to train our own large language models or third-party models, and we've stayed true to our word. However, AI is still a new field, so it's crucial to educate customers on what that means. Some customers might assume that we're using their data, but that's not the case. Even if some individuals want to opt-in to help us refine our data, we decline their offer. Overall, our priority is protecting our customers' data and ensuring they understand our practices.

Taking a responsible approach is crucial and one that we have already committed to. It is why our customers are drawn towards our AI Companion feature, with over 700,000 accounts already enabled. This number is only set to increase as more customers trust and choose to enable it. In particular, larger enterprise customers carry out an internal audit to ensure that none of their vendors turn on the AI feature and meet their specifications for internal security and privacy. Thankfully, they review our AI stack, policy, commitment, and terms of service, allowing them to feel comfortable and confident in their choice.

When you consider the significant plan - which still amazes me that this is what you aspire to achieve - of sending a digital double to a meeting to make decisions on your behalf that everyone trusts, agrees with, and acts upon, the risk of privacy violation is even greater. The area of security becomes more suitable for corruption. If someone can access my Zoom account and control my digital double, that's a critical issue. How do you plan on managing this risk as you work towards achieving this vision?

That's a great question. When it comes to privacy and security in virtual meetings, there are two important things to consider. Firstly, it's crucial to ensure that nobody else can gain unauthorized access to your meeting. This is particularly important when you're communicating with someone like Eric, and not a stranger. Secondly, it's important to make sure that your conversation remains confidential and secure during the call. We recently introduced the industry's first post-quantum encryption, which is a huge step forward in ensuring virtual meetings remain private and secure. Additionally, we're working hard to address the potential issues that may arise from deepfake technology. This requires much more than just two-factor authentication; we're utilizing AI to improve this experience and protect against potential deepfake threats.

When it comes to your project, you have to consider both sides. If your goal involves having a digital version of yourself attend a Zoom meeting and make decisions, then you must create a deepfake of yourself that looks and acts convincingly in those scenarios. On the other hand, you also need to be able to identify when the digital twin is being used. Which aspect of your project requires more attention and resources? Is it creating the digital twin or detecting its use?

However, which one is larger?

I believe that these two aspects are heavily intertwined. It's impossible to concentrate on just one and disregard the other. Doing so will not yield desirable results. Thus, we must provide a comprehensive experience from beginning to end and ensure that it's executed flawlessly. I also opine that it's crucial to allocate resources and attention to both components equally.

Do you believe that it will be feasible to accurately identify digital replicas or manipulated videos?

Definitely. Once more, it's like this: If I send an identical digital version of myself, it will be verified. You'll be able to confirm if it's genuinely the Eric digital twin, or belonging to someone else, due to cutting-edge technology and innovative innovations. I believe it's highly possible to identify. Otherwise, if you send a digital twin, it might not be me - you could end up meeting someone else entirely.

Do you ever wonder about the boundaries of what's possible? Imagine if I had a virtual version of myself that could attend online meetings and conferences - would it be acceptable for me to create 100 or even 1,000 copies of myself to attend different events simultaneously? Is there a point where this becomes excessive? What do you think about the idea of putting limits on this technology?

I believe that there is no set limit for the number of digital twins that a person can have. It completely relies on personal preference, similar to deciding whether to wear a black or white jacket to a meeting on a particular day. Ultimately, the digital twins belong to the individual and they can have as many versions as they desire. Personally, I may only require one version.

Is it possible for all digital replicas of oneself to be linked together? For example, if I have 100 digital twins scattered everywhere, and one of them asks if I want my next car to be painted red or blue, while another suggests white or black, how will they communicate and know about each other's responses?

First and foremost, it is important to note that you possess complete control over all your digital replicas.

There are two important points to consider. Firstly, your digital twins, which are basically multiple versions of yourself in the digital world, can vary greatly depending on your training and skills. For instance, one digital twin may excel in sales while another may have expertise in engineering. It’s up to you to manage these different digital versions of yourself. Secondly, whenever you decide to send one of your digital twins to attend a meeting or participate in a digital context, you can rest assured that they will be recognized as authentic through AI-based authentication. Everyone involved will know whether it’s a digital twin of Nilay or one of Eric’s digital twins.

Can you envision this occurring at Zoom's data center? When I access my Zoom account, I have my engineer and designer virtual copies, and potentially others, and instruct them to carry out tasks using the Zoom interface. Or, do I possess these virtual copies and establish a connection with Zoom?

In my opinion, the interface appears to be Zoom's. However, the way of managing it is quite distinct. This is why I am fond of crypto technology since it is more decentralized. As far as I know, it's not possible to store a copy of yourself on our server. Instead, you might prefer to save it somewhere that you feel secure, such as on your phone, desktop, or even where you store your Bitcoin. I don't believe it's necessary to give your digital twin to every vendor. Regardless of whether you use Zoom or other services, this isn't the architecture we follow.

I am very interested in this topic because constructing a substantial amount of this technology is essential for it to work effectively. This is the overall goal that we are working towards. A few of these inquiries seem to be quite significant. How fast can digital twins be implemented? This is a crucial decision that we must all collaboratively consider. Has it been considered that an individual may want to distribute a thousand digital twins at one time? This could potentially result in an unusual situation.

I completely agree with you. This is why AI is unpredictable, but it will definitely become a reality. I utilize my LLM aimed at consumers which trains with various digital replicas, and my friend shares the same faith in this too. It is undoubtedly effective. However, using technologies like distributing computing technology, AI technology, AR technology, and crypto technology can have some drawbacks. This is why the next 10 or 20 years will be even more exhilarating than the last two decades.

Moving back to the topic of Zoom, I was curious about what the company's future might hold, but ultimately it's up to them to make it happen. Something that caught my attention was that many of Zoom's employees have returned to working in an actual office. I've heard you mention in interviews that you experience "Zoom fatigue." Can you elaborate on that for me? It seems like a challenge to promote software for remote work and then ask your employees to come back to the office because they're tired of using it.

Let me clarify something for two reasons. Firstly, our company is dedicated to providing happiness to our customers by building better products. It's important to us to lead by example and use our own products ourselves. Many of our customers work in a hybrid capacity, spending some days in the office and some at home. In order to support this way of working, we create features that cater to this hybrid model. It doesn't make sense to not support it when we are promoting hybrid work. When our customers come to us, they often ask to see how we support hybrid work. They want to see the features in action and know how we use them ourselves. For example, if we have a feature that allows employees to reserve a desk, customers may ask us how we utilise that feature internally. If all our employees work from home for five days a week, I won't be able to show them the features in use. That's why it's important for us to embrace hybrid work ourselves.

The second reason why Zoom is different is due to the COVID-19 pandemic. They hired over 6,000 new employees during this time, which is quite different from companies who already have established teams. Because of this, building trust among new and old employees is a challenge. Zoom believes that in-person meetings can help build trust much faster than virtual meetings. This is a unique aspect of Zoom and is driven by their need to support hybrid work. They encourage their employees to work from home for five days a week, only coming into the office for two days.

I totally agree with your viewpoint. We have employed several individuals and I believe that being present physically is crucial whereas digital tools help in accomplishing your work. How is this reflecting in your workplace? Are employees gathering more frequently face-to-face? Are they keen on doing so more often? Are the digital tools contributing to this or creating more distance?

I believe that individuals aren't interested in meeting up often in person, but they crave getting together once in a while, particularly new hires. Initiating with face-to-face communication, we can build a strong relationship. Further updates can be shared online. In addition, we can have an online conversation after our first physical meeting, but it would be nice to come together in person annually as we have many staff living in LA, Florida, and Texas who would appreciate this opportunity. Overall, meeting up twice a year is a good enough standard.

Are you referring to the practice of using your own product for testing purposes? It's essential to have a hybrid workplace to develop effective tools for such workplaces. However, incorporating AI carries some uncertainties. We've discussed the potential risks and benefits that come with it. As we introduce more AI functionalities, we must also be prepared to take on some risks in Zoom's operations. Have you pondered about how much risk you're comfortable with in using an untested AI tool?

An excellent question! To ensure the safety and privacy of our customers, we have an internal process in place. Our team responsible for security and privacy also has a separate product team within them. This is because it's important for us to be responsible when announcing new AI features. Instead of immediately rolling them out to our customers, we first let our internal employees, the Zoomies, try them out and give feedback. We only release the features to our customers after enterprise IT tests them and feels confident using them. In addition, we take a responsible approach by making sure some AI features are turned off by default, and only enabled after testing.

This will definitely hinder the speed at which people start using those AI features, but we can try a different tactic for other clients. With certain accounts, we will enable specific features automatically. We analyze the situation from various perspectives and guarantee that everyone has all the necessary controls - enterprise, account IT. As the end user, you have control as well. The team also has control and we have an internal testing process. We are being extremely responsible in our approach.

Can you provide me with an instance of an AI function that you experimented with and had the thought, "Wow, this is fantastic. We should make it available to all members of the organization"?

I believe it was almost a year ago when we initially introduced the AI Companion feature's first aspect, namely the meeting summary. After testing it out, we came across some probable glitches, leading us to decide that it's best to hold off on releasing it to customers. Additionally, when we do make it available, it will be turned off by default. For this feature, enterprise accounts and IT must take charge as they have to be confident in it before passing it on to customers. That's essentially it.

Let me share another recent example with you. In a meeting with a client, we discussed a feature where AI is used to create a virtual background. With this feature, I can select a different virtual background for each meeting simply by typing keywords like "Happy Monday" or "Busy Friday". We tested it thoroughly before releasing it to ensure there were no potential privacy or security issues. We fixed any problems until we were completely confident that our customers would feel comfortable using it and not experience any problems. Now we're ready to release it.

Okay, let's explore the opposite viewpoint of that query. Can you name an AI characteristic that you may have created or seen someone else create, which you wouldn't be at ease utilizing in Zoom's operations?

Imagine I need to make a video. When I finish recording, I usually ask for someone's help or use a machine to say, “Can you send me a few minutes of video?” Later, I can send a written script and the video will be made by itself using technology. We are currently developing this feature, but I am not completely positive about it because it may create problems with privacy. We need to address this issue before proceeding.

Do you believe that this is truly related to the concept of digital twins, allowing for the creation of lifelike videos using an individual's face? What needs to happen before you feel confident using this technology yourself?

It's important to ensure that over 8,000 individuals are using Zoom and are comfortable with the platform. Additionally, we want to introduce early adopters to the new features and ensure they are at ease using them. Once we gather feedback and everyone is comfortable, we can make the features globally available. Before making any AI features available, we follow a conservative and responsible release process.

As we write about AI on The Verge, I've been reflecting on a noticeable difference in attitudes towards AI between companies and our audience. Companies are invested in AI, whereas our audience is mostly skeptical and sometimes even outraged. It's unclear whether the tech industry has acknowledged this divide.

Let me provide you with an illustration. I believe that the situation with Scarlett Johansson and OpenAI highlights the disparity. In addition, the Apple commercial in which they destroyed objects with an iPad, was not related to AI but was still disapproved by individuals who possess a general dislike towards AI. This is a prime example of the gap. What are your thoughts on this matter? We must now deal with how the common people perceive technology versus how the industry perceives technology. It appears that it is time to tackle this issue head on.

This is an excellent question. My perspective on this matter is that for any innovative technology, it's essential to consider human behavior. You need to concentrate on those who are willing to try new things first, improve the technology constantly, until it's readily accepted by everyone. As a new technology, you can't expect every individual, group, or business to adopt it from day one. It's not a reasonable assumption. Instead, focus on early adopters, and slowly expand from there. That's the key.

When it comes to new technology, it's normal to have new features introduced. However, the key to success is gradually improving the technology and ensuring that all users can benefit from it. This applies to AI too. It's why we feel confident about it, and I believe many others also share this sentiment, particularly individuals. We can't simply force users to adopt a new feature by saying it's cool. It takes time and gradual improvement to get everyone on board.

Eric, you've got a grand vision for where Zoom is headed. What's the upcoming milestone that people can expect to witness? What lies ahead for the future of Zoom?

I believe it's important for users to understand that Zoom isn't just a simple videoconferencing platform- it's a tool for collaborating in the workplace. One thing to note is that the Zoom AI Companion offers many fantastic features and is completely free to use. I am a firm believer that more users should take advantage of the various AI features the platform offers. While we've already rolled out numerous features, we plan on introducing even more AI options. I believe that as more customers utilize these new features, they will discover innovative ways to enhance their work experience.

Wow, Eric, that's fantastic! Many thanks for joining us on Decoder. We hope to see you again soon, maybe even as a virtual counterpart in the near future.

Alternatively, you could meet face-to-face.

Absolutely. Many thanks, pal. I genuinely value your gesture.