Glasswing officially launched in May of 2016, but Seseri has a long history of investing in companies offering AI solutions. Previously, she worked as a partner at Fairhaven Capital, where she funded AI-driven companies such as CrowdTwist, SocialFlow, and Jibo.
Below, Seseri talks with Zach Winn about investing in and starting AI companies.
Zach Winn: In what industries would you like to see more AI startups?
Rudina Seseri: While certain industries will naturally gravitate to including more, or different facets of AI, whether it’s vision or speech or different learning techniques, and there are characteristics of different datasets we can talk about, but I think we’ll see AI startups across the board.
Having said that, where AI is naturally planting itself in these early days is in markets that are data intensive or lend themselves to data, so we see this taking root already in enterprise areas like sales and marketing, IT, and HR.
We’re also seeing AI play a role in cybersecurity in a very meaningful way. Traditionally, enterprises have been on the receiving end and reactionary side of cybersecurity attacks, but with machine learning, we can predict that we’re going to be attacked and take corrective action in advance of an event. That’s a paradigm shift and a huge market opportunity.
And then lastly, we’re seeing a number of emerging AI startups in the field of robotics, whether it’s industrial robotics or consumer robotics. This is also true to self-driving cars, which are sensor heavy and produce a lot of data, which has driven investment in those categories. The next generation will surely produce a data-driven system of systems.
Outside of my firm’s focus areas, I think medicine and life sciences will be another key area for AI startups when you think about drug discovery, medical records, and tracking. These are data-heavy fields and all of them, in my view, lend themselves to AI.
ZW: Do you think the AI revolution will be driven by deep learning exclusively? In general, how big of a role will data play in AI’s growth?
RS: I think learning is woven across AI, whether it’s vision, speech, motion, or emotional cognition. Deep learning is one technique, but there are many others and the right one depends on the task at hand, the performance goals and the constraints you have. Today, deep learning is particularly strong, especially when it comes to vision, which is why you’ve seen Nvidia rise and Google put a lot of muscle behind deep learning.
You can’t have any kind of learning without data, so I think you need both models and data. The challenge we have in the AI paradigm, especially for deep learning right now, is that the required training data sets are very large, difficult to come by and typically task specific. But this is an active area of research and we are seeing new methods being developed to make models more data efficient, to create training data computationally, and to better generalize across tasks.
But to your question, will data remain important? For sure. Are there also huge data dependencies that are a barrier for startups? Absolutely. But in most applications, we’re still in the early days of AI adoption and the performance requirements of the market still require focused approaches. Besides that, the startup ecosystem, researchers, and the entire AI industry are working to reduce that data dependency, and we’re coming at it from both the data side and the algorithm side.
ZW: How should startups be thinking differently about creating an AI company compared to a company using other tech?
RS: When you’re selling to Fortune 1000 companies, there’s no doubt AI is a focus topic at the C-level, but people also don’t know how to wrap their heads around it. And then it’s also true that leading in with AI as a startup can be problematic.
From the CEO on down, executives understand that AI is the future and they need to get up to speed and start to implement the technology. But then a company comes in and pitches to you, and people feel uncomfortable buying your product because they can’t fully comprehend how it works and how they need to adapt their expectations and how they work to implement a system that requires data to learn and has ever-increasing performance. So, what I tell my portfolio companies is to lead with the problems they solve.
If AI is integral to your tech, you as a CEO need to be able to articulate what it does, how it does it and why it’s differentiated. Startups need to be really thoughtful about that, and they need to be able to explain the outcomes, methodologies, what they test for, even if it’s at a high level. I don’t need the CEO to be an AI guy with a data science background, but if you’re pitching for funding or, more importantly, to a prospective customer, and you, as the CEO, can’t articulate your solution, it tells people you don’t know what you’re doing and you’re just jumping on the AI bandwagon.
ZW: What kinds of founding teams do you looking for in AI startups?
RS: The CEO can be a business person or a technical person but they need to have the capacity to build companies and not just products. Typically, someone has to have deep understanding of data science, computer science and the business problem they are tackling. And this can be all the same person or different ones. However, no two startups come in the same shape, which keeps things interesting. But company culture and people are so important. I say this over and over: People make or break you. I pay great attention to the team dynamics of these AI startups.
ZW: What are some advantages startups have over existing companies when trying to bring AI to a new space?
RS: Startups are AI natives. They get built from day one with the latest developments in a field that is changing at lightning speed. And if new companies know how to get the right data sets up front, it’s really hard for incumbents to catch up to that.
And then there are all the usual advantages a startup has over existing companies, like agility on execution and being unencumbered.
ZW: What are some disadvantages startups have when competing with existing companies?
RS: The data problem (when relevant), then the balance sheet, and the need to compete with the big players for talent.
ZW: Are there any specific spaces where you think startups have a particularly large advantage over existing companies in bringing AI solutions to market?
RS: We’re certainly seeing it in security market and in markets where there’s been some software to date but that have been traditional services businesses, typically supported by large groups of people, with a consulting element. Talla [the first Glasswing investment] is a perfect example, where it’s less about replacing workers and more about augmentation.
Then I think there are areas that have a clear need but the tools and services are hard to build and not enough companies have done it. For instance, what will be the AutoCADs of the world for specific AI tasks like vision and speech and other areas where you’ve seen green space and now companies are being created?
ZW: How does Glasswing differentiate itself from other VC firms with AI strategies?
RS: VC firms fall into two categories: One type is the generalist firm that says they do tech and part of that is AI. These firms are just coming up the learning curve. The other category is the firm that has a more focused strategy analogous to ours, and who are trying to get deeper into AI.
The way we’ve differentiated ourselves over time is we have over a decade of investing experience. So when we’re getting a pitch, as an example, and we get this comment all the time, they’ll say, “In the first 45 minutes with other VCs, we’re spending half our meeting educating them on what exactly AI is and what it means. But you guys are able to explain the rationale for why we picked approach A versus approach B.” That gives us an immediate bond with entrepreneurs based on our deep level of understanding and domain expertise.
We also have a large set of advisors - renowned entrepreneurs and technologists, AI visionaries, and world-leading executives that exclusively advise and support the firm and our portfolio companies. A large subset of those people specializes in AI and machine learning. So we’re helping our portfolio companies not only by understanding AI, not only by helping them understand the end market to invest in — I am focused on enterprise, Rick Grinnell (Managing Partner) does cybersecurity and robotics and Sarah Fay (Managing Director) does marketing and digital — where we have end-market expertise, AI expertise, and we extend our footprint between our relationships with advisors, CxOs of other companies, entrepreneurs, so it’s a big group that works exclusively with Glasswing and our portfolio companies.