Speakers:
- HubSpot: Yamini Rangan, CEO
- Zoom: Eric Yuan, CEO & Co-founder
- Fellows Fund: Alex Ren, Founding Partner (Moderator)
Summary:
The fireside chat featured Yamini Rangan, CEO of HubSpot, and Eric Yuan, founder, and CEO of Zoom, moderated by Alex Ren, founding partner at Fellows Fund, discussing the AI revolution's impact and the importance of embracing AI in business. They highlighted advancements in generative AI, customer interest in leveraging AI for productivity, and the integration of AI into their companies' platforms. Both emphasized responsible AI development, customer trust, and the need for innovation. The chat covered AI regulations, the balance between cost reduction and AI investment, and opportunities for startups in the AI space. The audience questions addressed real-time conversation modification and investing challenges, emphasizing the need for an open-minded approach. Overall, the chat provided insights into the AI revolution's current state and implications for businesses.
Video:
The full transcripts:
Alex Ren:
Alright. So, I'm Alex Ren, the founding partner of the Fellows Fund. Just a brief introduction to the Fellows Fund. We are an early-stage AI venture capital firm, built by 25 AI executives from top tech companies like Google, Facebook, Microsoft, and other top tech companies. It's my honor to moderate this fireside chat, featuring two great leaders in the space: Yamini Rangan and Eric Yuan. We do share one thing in common. We are all first-generation immigrants to the United States. Let's give us a big applause. Thank you. I know we all started with humble beginnings. I remember Eric tried nine times to apply for a US visa. He eventually came here and founded Zoom, which has benefited millions of people during and after the pandemic. Yamini's story is equally inspiring. I remember she came to the US at 25 with just a few hundred dollars. Her first job was in a restaurant. Eventually, she became a leader in top tech companies like SAP, Workday, Dropbox, and finally HubSpot. Leaders like them are truly our role models. They embody the spirit of entrepreneurship in their blood. Let's dive into our conversation. We're also going to leave some time for the audience to ask questions. Let me start with my first question. Seeing the outbreak of generative AI development, what's your personal feeling? Does it rekindle your passion to work in technology?
Yamini Rangan:
Okay, I can get started. Eric, nine times? I didn't even know what happened nine times. This is a story that we have to catch up on. But it is so fantastic to be here and see the energy. This is the entrepreneurial energy of the valley. It's a fascinating time and such a great time to be in technology. To answer your question, Alex, absolutely. I've replaced my Netflix binge-watching with Gen AI podcast-watching. There's so much happening right now in the industry that it feels like things are moving at the speed of light. Every week, there are 10 announcements and 10 new significant progressions in the field. I started my career as an engineer, did my undergrad in India, came here for a master's, and one of my first subjects here was neural networks. In the early 90s, neural networks were just getting started. To see the level of progress in what neural networks are able to do, which is really the foundation for any of the large language models, is just fascinating. So, it's great to be here and in technology at this time.
Eric Yuan:
Thank you. When you ask this question, my feeling is — I feel very old. In the early days of the Internet, I spent a lot of time trying to understand how it works and familiarize myself with the Internet Protocol, HTML, routers, and everything. I felt like I could catch up. Now, if you look at what's going on in the generative AI world, there are so many research papers, large language models, and companies, as well as a lot of acronyms. I can't remember any of them anymore. But this signifies the innovation speed in the AI industry, particularly in generative AI. It's moving rapidly, providing huge opportunities. I am the CEO of Zoom, otherwise, I would likely start a company to focus on AI. When you look at the technology stack, there are many areas to innovate, such as making large language models more scalable, or finding better models than transformers. How to apply generative AI to all those vertical use cases is a huge opportunity. This is the best time for any of you interested in AI to start a company.
Alex Ren:
I disagree with Eric; entrepreneurs can never be old. So, after the launch of GPT-3, did you receive customer calls for more AI integration in your products? Has AI become a top priority for your respective companies since then?
Eric Yuan:
Before GPT-3 became popular, we had already heavily invested in AI. Features like virtual background, noise suppression, and hand gesture recognition, all have AI involved. This is why we heavily invested in AI resources over the past several years. Whenever I talk with customers, they ask about how we leverage AI to improve our product experience. The second question is always, "What's your AI strategy?" It's very important. Every business has to embrace AI, otherwise, they are going to lag behind. At Zoom, we ask every product manager and engineer how they can leverage AI to improve features or the product roadmap.
Yamini Rangan:
I completely agree. There are two parts to your question: Are customers asking, and what's the conversation you're having with customers? And has AI become a priority within the company? When I talk to customers, they typically fall into three categories. The first one is early adopters who are already experimenting with AI and looking for ways to improve their own effectiveness and productivity. The second group is interested and learning but doesn't quite know exactly how to apply it beyond the hype. The third group is waiting for us, and they're interested in leveraging generative AI, but they don't have the time to go through the hype and learn about the 20 models that we just talked about. The technology is developing at a much faster rate than customer adoption of the technology. From a HubSpot perspective, this year we've really focused our whole organization on generative AI and leveraging its use cases. So, has AI shifted company priority? Absolutely. You can talk to almost any product manager, and they are looking at what we can do with generative AI to drive effectiveness for our customers.
Alex Ren:
Thank you very much. Let's talk about the specific use cases. How do Zoom and HubSpot use AI today? What's the impact on your customers?
Yamini Rangan:
I'm very keen to hear about it because I'm always on the lookout for the next set of features from Zoom. From HubSpot's perspective, we mainly serve small and medium businesses that generally don't have AI or large language model experts. Our goal has always been to democratize complex technology for SMBs. We aim to help our customers generate content, insights, and code faster. As part of our product strategy, we're integrating AI into the foundation of HubSpot, not just as a feature of one product, but at the platform level. We're exploring ways to leverage large language models.
In March, we launched a feature called ChatSpot, a natural language interface to HubSpot. It simplifies the CRM experience. Instead of needing to be a marketing or sales expert to navigate reports, users can use the chatbot to ask questions like, "What was the web traffic like? How did it compare to last year? Can you create that report and send it to me every Monday?" This type of natural language interface is driving user engagement. In the last two months, we've had over 40,000 users interacting with it, providing feedback, and helping us improve.
Another Alpha stage feature is content ideation and creation. We believe this will become essential in the future. It allows marketers, salespeople, or service personnel to generate or summarize content. Although it's early days, we're excited about the potential value this could bring to our customers.
Eric Yuan:
At Zoom, we already have several AI features in beta. These include a meeting summary feature and real-time summary generation for latecomers to ensure they don't miss anything. We also have a team chatter solution comparable to Slack. If you haven't tried this feature yet, I recommend you do. It works well and it's free. Last year, we announced an email and calendar feature that uses AI to assist with writing emails. Our whiteboard will soon integrate with AI, and our contact center virtual agent already incorporates various AI features. There are a lot of exciting developments in the pipeline.
Alex Ren:
Thank you. I use these features daily. My last question today is about open-source vs. closed-source language models. I know you're both working with language models similar to OpenAI's GPT. Are you considering open-source models like Llama or Dolly?
Eric Yuan:
We're open to all possibilities from a user perspective. We spend a lot of time communicating with our customers to understand their needs and the issues they face. We've adopted a federated AI approach. Not only do we have our own large language model, but we also embrace others. We've announced collaborations with OpenAI and recently invested in Anthropic, a company specializing in language models. We also support open-source models. Some large enterprise customers have told us that they have their own large language models and do not want to use any model we offer. We support that too, by integrating our customers' language models. In short, whether it's open source or closed source, our customers’ model or our model, we support them all. We center our approach around customer demand and this necessitates a flexible AI approach.
Yamini Rangan:
I wholeheartedly agree. From our perspective at HubSpot, we are open to multiple large language models (LLMs). We're currently collaborating with OpenAI, but open source is catching up rapidly, making significant improvements in just a few weeks. Like Eric and Zoom, we prioritize our customer's needs, which will likely lead us to work with multiple LLMs. The pace of development in this space is exciting.
Alex Ren:
Those are insightful answers. A crucial factor in training a good model is high-quality data. Could you share more about your data strategy, such as whether you have any proprietary data to train or fine-tune models?
Yamini Rangan:
That's a pertinent question. Currently, we are working with several large language models, and I distinguish between a 'data out' strategy and a 'data in' strategy. By 'data out,' I mean that if our customers, using HubSpot and interacting with LLMs, prefer not to contribute their data to training LLMs, they have full control to opt out. It's paramount to us that control over data rests with our customers.
Regarding 'data in,' it pertains to data within HubSpot. Our focus has always been on empowering our customers, allowing them to decide what they can and cannot share. Transparency is key. We want our customers to understand how we use their data, and we provide clarity about their controls. There's a wealth of data available; the challenge is how to utilize it both incoming and outgoing.
Eric Yuan:
I agree completely. A very well-putted answer.
Alex Ren:
Thank you. In recent times, there have been congressional hearings featuring OpenAI's CEO and other AI leaders discussing potential AI regulations. What do you think could be the implications for the enterprise software industry if we see increased AI regulations?
Eric Yuan:
I believe Sam did an excellent job at the hearing. We all owe him our gratitude. The key takeaway is that every company must adopt a responsible approach to AI, prioritizing safety and ethics. If we fail in this regard, we're likely to lose the trust of our users. At Zoom, by default, we do not use customer data to train our models; we always seek customer consent. If we neglect this aspect, no matter how great our AI model might be, customers won't trust us. Once trust is lost, it's hard to regain it.
Yamini Rangan:
I couldn't agree more. If you haven't listened to those congressional hearings, I strongly recommend you do. Two points struck me during those conversations. First, there was a bipartisan agreement to solve the problem together, which is not common. Second, industry leaders were asking for regulations, which is also unusual. Transparency and customer trust were central themes. As an industry, we need to work with educational institutions and policymakers to make AI beneficial for humanity. I find these developments quite encouraging.
Alex Ren:
We'll have to see how it evolves. Moving on, as public companies, how do you balance cost reduction and investment in AI? Also, how do you think AI will reshape the workforce?
Eric Yuan:
I believe that AI is set to change everything on many fronts. From our perspective, we need to consider it from a product, process, and people viewpoint. When it comes to products, we've been discussing how crucial it is to embrace AI and leverage it more and more to improve the product experience. Internally, we also need to look at how we can use AI to enhance the process workflow. That includes fully automating the process, identifying any weak points, and refining it accordingly. Finally, improving productivity through AI is vital in all areas. In order to excel on the product side, we must embrace AI on various fronts, since it will revolutionize everything we do.
Yamini Rangan:
Indeed, I hold the view that AI is not going to replace humans. However, humans who utilize AI are likely to replace those who don't. We are in the business of helping people leverage technology and democratize it. When we consider productivity, it's clear that some repetitive, low-risk tasks will get automated. That means the value that humans can add will actually increase, not decrease. We examine both products and people in this light. How can we help marketers become even more effective, not just by generating large volumes of content, but high-value content? How can we aid salespeople in forging stronger connections with customers? This is the lens through which we are approaching this issue.
Alex Ren:
Yes, much like previous technological revolutions, technology always benefits us and brings more opportunities. I believe in that as well. We have many audience members who are founders and investors in this space. What opportunities do these individuals have in this AI revolution, and how can small startups compete given that big companies control numerous models and vast amounts of data?
Eric Yuan:
I think, as I mentioned earlier, there's a massive opportunity here. If you're an entrepreneur thinking of starting a company, I'd say this is the best time. Don't assume that large companies can innovate faster. Take OpenAI for instance. Five years ago, nobody had heard of them. They were a small startup, and look at them now. The same goes for other successful startups. A friend of mine recently asked me if I'd tried OpenAI, confusing it with a different company. Many of the cool technologies we see today aren't developed by large companies with thousands of engineers but by nimble startups. I see a massive opportunity for startups because they can move quickly and innovate faster. That's why when something new, like general AI, comes around, it's the best time for startups as they are very agile.
Yamini Rangan:
I would absolutely second that. While incumbents have some advantages, there are also several benefits for startups. If you think about incumbents, data and how they use that data is a competitive moat. They also have the ability to get feedback from customers and iterate much faster than others, as OpenAI did, which helped them gain millions of users in a short period of time. On the other hand, this is a time when the processes that have powered knowledge workers can be reimagined. Entrepreneurs and startups have the ability to be more nimble, start with a clean slate, and rethink what work needs to be done and how it can be powered. So, it's a great time to be both a more established technology player as well as a startup. I think we're going to have a lot of fun over the next decade with this.
Alex Ren:
Yes, that's a great answer. I totally agree with you. As an investor, I often use the analogy of the internet during the early 90s. You know, Yahoo opened up this new space. Similarly, today, new spaces are being opened up, but there's a huge opportunity for startups and other companies to apply AI to different verticals. That's why I deeply agree with you.
Eric Yuan:
So Alex, if I understand correctly, the implication of your question is that if someone has a good idea and wants to start a company, make sure to talk with Alex, they have a Fellows Fund and can invest in you guys.
Alex Ren:
Yes, exactly. In fact, we've already invested in 10 companies in the AI space in the last two months. So, I totally agree with you. We have a little bit of time to take maybe two more questions. Please state your name, company, and question.
Audience Question 1:
It's wonderful to be here. Thanks, Alex, for co-hosting us. My question is for Eric. It's an ethically fraught question but an important one. The modification of what people say in conversation via Zoom is an application that my friends and I are all fascinated by. I'm the CEO of a company called Omniscience. We build retrieval-augmented generative systems so that all of your data, everything you've ever written or seen or talked about with people, is behind a larger language model. I want to use this model in my Zoom calls to augment my decisions and to answer my questions in ways that are similar to what I would have said, but that are much more detailed and well thought out. I believe every employee that I have can perform much more effectively in the face of that kind of feature. I wonder if you think it's too ethically fraught to build it, or if that is Zoom's intention.
Eric Yuan:
We all generate a significant amount of data in our conversations, right? Either you record a meeting or it is transcribed. That data is valuable. In fact, a summarizing feature will be available very soon on Zoom. Not only do we support this, but we also plan to support plugins in the future, so you can apply your own models. Of course, we try to make all the data available to the meeting host, so you can apply your own model or leverage our own model. Essentially, we're going to have this feature. Yes, it will be in real-time, and also post-meeting as well. For example, after a one-hour meeting, a summary will be available. Plus, we also plan to expose an API for you to add your own plugins.
Alex Ren:
Great. Next question, please.
Audience Question 2:
This is Scott Rose from Flying Ventures. One of the things that keeps me up at night is making the wrong investment. In the world of generative AI, it looks like there are a lot of opportunities where companies are placing a bet on a fine-tuned foundational model that belongs to somebody else. What is your opinion with regard to the risk, given that, as you said, Congress is about to regulate the industry?
Eric Yuan:
That's a tough question. I don't have a great answer for that. However, I will say that given all these opportunities, it's hard to bet on just one company or one technology. You have to be open-minded. It's like the early days of the internet. You have to take an open-minded approach as an investor, and probably consider a lot of companies, regardless of the challenges they're facing. You never know who's going to win. That's why our approach is always to take an open-minded approach, adopt all sorts of AI models, talk with every startup company, and invest in as many potentially successful companies as we can.
Yamini Rangan:
Your job is exceptionally difficult. When you're dealing with a big technology transition like this, you have to look at every layer of the stack, who is getting disintermediated, who's got competitive moats, and who's going to survive. It's hard to predict. If you had asked six months ago, you might have thought there would only be a few large language models. Now, you might predict there will be many specialized models. Every layer of the tech stack is going to be transformed as we figure this out. During these early stages, you have to form a hypothesis, make a few bets, and know that you might have to adapt and shift quickly. At HubSpot, we're trying to understand who has the advantage and what that advantage is. We spend a lot of time talking to AI experts from academia and the industry. There's no clarity on where we'll be three or five years from now, but there is clarity on what we can do right now: get a ton of feedback, iterate, and really identify use cases that are going to be table stakes. You also have to identify use cases that are going to be specialized based on the data that you have. It's an exciting time but also a time of transition.
Eric Yuan:
When most people look back at the search engine landscape in the 1995 to 1996 timeframe, they realize there were so many search engines. It was indeed tough to decide where to invest. Guess what? Eventually, Google, which was born in 1999, dominated the space. If you had lost focus then and thought search engines were over, you would have missed out on a great investment opportunity. I think the same principle applies to today's AI world. You have to be patient and open-minded. Always keep an eye on promising startup companies. Otherwise, you might miss out on the eventual winner.
Alex Ren:
On that note, thank you very much for your insights. Let's give a big round of applause. Thank you, Eric and Yamini.
Call to Action
AI enthusiasts, entrepreneurs, and founders are encouraged to get involved in future discussions by reaching out to Fellows Fund at ai@fellows.fund. Whether you want to attend our conversations or engage in AI startups and investments, don't hesitate to connect with us. We look forward to hearing from you and exploring the exciting world of AI together.
