Speakers:
- Zoom: Smita Hashim, Chief Product Officer
- Zoom: Mahesh Ram, Head of Digital Customer Experience
Summary:
In a panel discussion moderated by Mahesh Ram, Head of Digital Customer Experience of Zoom, Smita Hashim, Chief Product Officer of Zoom, who recently joined Zoom after an extensive career in product management at Microsoft and Google, shared her background and what attracted her to Zoom. She highlighted Zoom's mission-based approach, belief in hybrid work, and the company's focus on video communications and solving unsolved problems in the space. The discussion then shifted to Zoom's strategic positioning in the current AI environment and its use of generative AI. Smita mentioned that Zoom has a wide product surface area and sees an opportunity to enhance user productivity and collaboration through AI features across various products. She highlighted upcoming features like meeting transcription and chat summarization and emphasized the importance of learning by doing and being responsible with customer data.
The conversation continued with insights from customer conversations regarding AI and generative AI. Customers expressed eagerness to learn more about Zoom's AI strategy and its support in their AI-related endeavors. They also emphasized the importance of data handling and expected prompt delivery of usable AI products. The panelists discussed Zoom's federated approach to AI, which involves leveraging multiple models, including partnerships with OpenAI and Anthropic, as well as Zoom's own proprietary models. They mentioned specific use cases where different models are applied, such as meeting summaries and chat compose. The panelists stressed the need for customization, transparency, and responsible usage of AI models. They also highlighted the importance of context in AI applications and optimizing models based on the specific use case to create user value. The discussion concluded with Smita expressing her excitement for the future of AI at Zoom, focusing on moving from hype to reality, delivering a spectrum of evolving capabilities, empowering users, and ensuring value from investments.
Video:
Full Transcripts:
Mahesh Ram:
Okay, I first just want to say welcome everyone. And I want to specially thank Fellows Fund and HubSpot for allowing me to channel my inner opera today as a moderator, with Smita. It's a rare privilege for me, but I'm looking forward to it. And I'll start with a basic question. You recently joined Zoom after over 20 years in product management with distinguished tenures at Microsoft and Google. Could you tell us a little bit more about your background? And more importantly, what attracted you to Zoom?
Smita Hashim:
Yeah, absolutely. Hello, everyone. It's great to be here. What a wonderful day of discussions. So I joined Zoom around 100 days ago, coming from Microsoft Teams, where I was leading the real-time communications and video-related products. Prior to that, I had a long tenure at Google, where I worked on various productivity products such as Google Meet, Google Voice, Calendar, Tasks, and Chromebooks. I've had a very interesting career, and what really attracted me to Zoom was my belief in hybrid work and the power of video communications. I think there are many unsolved problems in this space that we should collectively address for the benefit of all. Zoom is a mission-based company that moves incredibly fast and offers the best video quality. It's also investing in internal and external collaboration. So far, it has been great to be a part of Zoom.
Mahesh Ram:
We are very excited to have you with us. AI and generative AI are everywhere these days, making headlines and hard to escape even on the weekends. How are you strategically positioning Zoom in this current environment with all these advances? And could you tell us a little bit about what Zoom is already doing with AI today?
Smita Hashim:
Absolutely. I agree that generative AI is pervasive and it's an incredibly exciting time. From Zoom's perspective, we have a wide product surface area that users spend a lot of time on. This is important because users tend to use tools where they spend their time. It provides us with a great opportunity to make users even more productive and enhance their collaboration experience. All our product teams, including Zoom Meetings, Zoom Chat, Zoom Phone, Zoom Whiteboard, Zoom Contact Center, and Zoom IQ for Sales, are actively working on building more AI features. We have been making significant progress in the area of generative AI, and we are rapidly bringing new features to the market. Some of these features will be generally available in the next couple of weeks, such as meeting transcription and chat summarization. We are moving swiftly because we believe in learning by doing. Additionally, we recognize the evolving landscape of AI models and the diversity within it. To address this, we are taking a federated approach to AI, which involves leveraging multiple models from the outset. It's crucial for us to be responsible with customer data and treat it with the utmost care, and that's an integral part of our product strategy.
Mahesh Ram:
As a founder coming into Zoom, I've noticed that Eric, our CEO, listens remarkably well to customers. It's truly unbelievable. In your first 100 days at Zoom, you have also immersed yourself in conversations with customers. What are you hearing from them about what they want from strategic providers like Zoom in the realm of AI and generative AI? And how did this lead to the formulation of our AI principles? Could you elaborate on that?
Smita Hashim:
Absolutely. So, yeah, I think one thing, again, Zoom has this fantastic customer base, ranging from large enterprises to small businesses. I have had extensive conversations with customers, including our largest ones, and the topic of AI always comes up. Customers are eager to learn and do more in this space. They want to understand our AI strategy, particularly with generative AI, and how we can assist them. They also express the importance of data handling and expect us to be responsible stewards of their data. These are the top concerns on their minds. Additionally, customers are eager to get their hands on AI products and start using them. They appreciate a well-defined strategy, guidance on data understanding, navigating the landscape of models, and prompt delivery of usable products.
Mahesh Ram:
That's a great answer. I would just like to add that in the last 100 days, I have noticed that the questions customers ask about AI have become more sophisticated, keeping pace with the increasing complexity of AI itself. For founders engaging with companies, be prepared for FAQs and continuous engagement.
Smita Hashim:
The conversations are incredibly engaging and energizing. We are fortunate to be working in this field.
Mahesh Ram:
Moving on to the federated approach to AI that Eric mentioned this morning. How will Zoom strike a balance between using third-party foundational models like OpenAI and Anthropic, and developing our own models to improve the Zoom experience? How will this determination be made?
Smita Hashim:
The landscape is evolving rapidly, and it's clear that we need to work with multiple models and optimize our approach at Zoom. We have already announced partnerships with OpenAI and Anthropic, which we are excited about. We also have our own proprietary models. Some of our larger customers express their desire to use their own models, and we fully support that. This is the essence of our federated approach. In terms of learning, optimizing, and applying these models, let me provide some context by sharing examples of our upcoming features. We are about to launch a Zoom Meeting Summary feature that will enhance efficiency for users. We are using our own proprietary model for this feature, which understands conversational data and multiperson conversations. We have a strong foundation in conversational intelligence through our product Zoom IQ for Sales, and we are applying that knowledge to Meetings. Additionally, we have the Team Chat Compose feature, a general-purpose conversational interface in Zoom Team Chat, where we are currently using OpenAI. We recently announced a partnership with Anthropic, and we are excited about applying their ontology to our Zoom Contact Center and Zoom Virtual Agent products, which have a rule-based nature. We believe in the power of constitutional AI and aim for more definitive answers. For example, our Zoom Virtual Agent, an unattended agent, requires high-quality responses and employs specific models. These examples demonstrate how we apply different models within the Zoom platform to address specific use cases. As we continue to evolve and learn, we will determine the next steps in our journey.
Mahesh Ram:
Absolutely, I would like to add that understanding the context in which someone uses AI or generative AI is crucial. By bringing that context into AI applications, we can differentiate ourselves from more generic approaches. The application context is still significant, as users engage with the application, they realize the need to fine-tune models and overcome challenges. Factors such as data sequencing and context provision play a role. For instance, our meeting summary feature, trained on Zoom IQ for Sales, needed fine-tuning for general Meetings. Similarly, Team Chat Compose with Zoom IQ requires understanding user tones and speech patterns, which becomes evident during application. Optimizing models based on application context is where user value is created.
Smita Hashim:
From a business perspective, which I have gathered from our customer conversations, businesses seek increased productivity and better outcomes while maintaining excellent customer and end-user experiences. They want their employees to be more productive, allowing time for creativity, collaboration, and relationship building, ultimately leading to improved outcomes. In customer-facing applications, businesses aim to optimize and leverage AI while ensuring that customer experiences remain uncompromised. Additionally, they are curious about the cost implications of these products, as free trials are prevalent, but monetization strategies need to be considered in the long term.
Mahesh Ram:
I would like to add insight from a recent customer conversation. They had a realization that core capabilities, such as summarization, can be applied across multiple business use cases. Summarizing meetings, extracting next steps, and understanding the meeting context are valuable features. We can extend this capability to the Contact Center, summarizing conversations between bots and humans, and synthesizing information for agents to address customer issues. The customer recognized our investment in these capabilities and the potential to apply them across various scenarios. It's about orchestrating features across the platform and delivering value. Now, looking ahead, what are you most excited about in terms of AI and how Zoom approaches it? What are your thoughts on the future of generative AI?
Smita Hashim:
I believe that AI applications will evolve rapidly, and my focus is on moving from hype to reality. I'm excited to see the features on our rich roadmap being built and fine-tuned, improving the user experience and gathering feedback from customers. We aim to progress from single product flows to orchestration across multiple products, expanding our assistive capabilities. Currently, users are in the loop, receiving information and taking action, but we aspire to become more proactive and offer a spectrum of evolving capabilities. Empowering users with increasing power and control is a key aspect. Additionally, I'm eager to witness the evolution of models, cost optimization, and monetization strategies to ensure value from our investments.
Mahesh Ram:
Indeed, there is tremendous excitement surrounding AI, but there are also concerns. Data privacy and security are paramount, and as application providers, we have a responsibility to be transparent and careful with customer data. Transparency and flexibility are essential, ensuring customers have control over their data and the ability to fine-tune models according to their specific needs, even for small customers. We must enable capabilities that allow customers to customize and train models while maintaining data control. Furthermore, it's crucial to address the limitations of third-party large language models and educate users about this. Responsible usage and highlighting the limitations of these products are vital to avoid unexpected outcomes for our customers.
Smita Hashim:
Absolutely, as you mentioned, transparency is crucial when it comes to customer data. We must provide customers with choices and flexibility, allowing them to use their own models and fine-tune them to their specific requirements while maintaining data control. Additionally, we need to educate users about the limitations of large language models and ensure responsible usage. It's important to highlight the potential errors and limitations to avoid misleading users. We have an obligation to provide a clear understanding of the capabilities and potential pitfalls of AI.
Mahesh Ram:
Indeed, even within our own company, we experienced the challenges and complexities of AI implementation. For example, when we brought GPT into our employee base, we had to consider issues of inclusion and accessibility. We faced questions about usage restrictions based on location and how to handle employees traveling across different regions. It's crucial to think about these complexities and ensure that we address them on behalf of our customers. By taking these factors into account, we can fulfill our obligation to provide a reliable and inclusive AI experience.
Smita Hashim:
Exactly. If I were to offer advice to young companies and startup founders in different lines of business who may want to partner with Zoom, I would emphasize the importance of being a practitioner. It's crucial to accelerate investments in AI, activate them, and bring them to market rapidly while ensuring quality and effectively communicating the benefits and limitations to users. It's also vital to understand the rapidly changing landscape and incorporate flexibility into application design from the beginning. Consider the user experience, backend flexibility, and cost optimization, including monetization strategies. For service and infrastructure providers, there is a vast opportunity to develop tools for foundational models or ML ops, evolving them and partnering with application providers to test and bring them to market.
Mahesh Ram:
In the next 100 days, what are you most looking forward to? What do you see on the horizon?
Smita Hashim:
These past 100 days have been incredibly busy, but also exciting. We have a rich roadmap, and we are rapidly bringing features to market. I anticipate that in the next 100 days, we will make significant progress. Customers will be using more of these features, and we will have actual products ready to offer. I'm particularly excited about the assistive capabilities we are developing, both for internal and external collaboration. For internal collaboration, such as meetings, phones, and whiteboards, we aim to create transformative experiences. For external collaboration, with products like the contact center, sales, and marketing, we have a strong foundation, and integrating AI capabilities will bring about truly transformative experiences.
Mahesh Ram:
Additionally, you've mentioned the concept of freeing up time for knowledge workers. Can you elaborate on your ideas for the future?
Smita Hashim:
Absolutely. As a collaboration company, we should strive to make the collaboration experience more efficient. One aspect I'm focused on is making meeting content easily digestible and actionable. We want to provide efficient ways to understand what happened in a meeting and follow up on it. Our goal is to save time by eliminating the need to read through extensive recordings or meeting notes. With the hybrid workforce becoming more prevalent, this will truly change the game for many of us. It's an area I'm personally very excited about.
Mahesh Ram:
Thank you for this insightful discussion. I appreciate your time and engaging dialogue. I also want to thank everyone in attendance, and I look forward to connecting with all of you. Thank you.
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.
