Listen on Apple Podcasts or Spotify

To kick off the panel on next-gen Generative-AI, moderator Sriram Ramakirshnan asked panelists their thoughts on the current climate for fundraising, and how it differs between GenAI vs. non AI focused companies.

HR: Raza shared that when Tavus was raising a seed round in 2020, investing in AI required convincing. In his words, “investors were taking a leap of faith”. This dynamic, of course, has changed in the last few years.

AP: Paten echoed Raza’s sentiment, pointing out that the major shift in fundraising is that where the use of ML models is not novel, the premium placed on AI in the last few years has changed the funding landscape.

VM: As for the frenzy around AI, Mugunthan credited the ecosystem in San Francisco for accelerating the access to resources (funding and otherwise) for Dynamo AI’s beginnings.

LC: Chang offered her insight as an investor, defending NYC as a hotbed for AI innovation and also highlighting the cyclical pattern that AI is tracking. Chang’s perspective on the landscape of the “AI frenzy” underscores the importance of diligence. She noted that with the hype around artificial intelligence, there will be a “road littered with carcasses of companies” which hastily applied artificial intelligence for the sake of doing so, rather than in solutions which would thoughtfully create a benefit.

Ramakrishnan then asked the panelists to share their thoughts on sales cycles in enterprise software, and how they approach resistance in the market when framing their products.

HR: When approaching enterprise sales, Raza stressed the importance of “meeting companies where they are”. There is intentionality around filtering for the right customers and the right timing, all the while being upfront about the risks. From a strategy perspective, Raza also suggests building trust by applying solutions in lower-risk scenarios (internal facing applications), finding a champion in partner organizations to build toward the future.

AP: Paten noted the start-up industry idiom “Move fast and break things”, warning that this is, in fact, the “exact opposite” of enterprise businesses. There is risk in introducing new products too quickly, and founders should be methodical about meeting enterprises where they are and picking the right lanes for product adoption. Paten offered an example in the context of insurance, noting that underwriting would be a weak lane to bet on in the insurance industry due to the widespread apprehension (despite the large opportunity).

VM: Mugunthan noted that in the first year of Dynamo AI, they did not sell to a single customer. However, the momentum of having one or two big customers creates strong tailwinds. As for industries where Mugunthan has seen more resistance and adoption, he characterizes verticals which are highly regulatory, particularly healthcare.

LC: Chang aligned with Mugunthan’s point, and acknowledged the “many flavors of enterprise sales cycles”. The hardest part about selling to large enterprise customers? Chang pointed out that with regulated industries, the challenge is knowing where in the lifecycle of the sales journey that you are.

When asked to share their thoughts on AI more broadly, the panelists touched on the commoditization of foundational models, open vs. closed source, their fears, and the future of mass-adoption.

HR: Raza set the table for his response in highlighting the value in domain specific companies building at the application layer. He believes that enterprises are open to buying solutions that feel truly tailor made to them. Raza’s company, Tavus, is a foundational model company that is powering the foundational layer, upon which such customized solutions can be constructed. As for his own fears about open source foundational models, and the commoditization of such models, Raza countered that naturally, most of tech becomes commoditized. He also notes the power of open source improvements to accelerate innovation in closed source, and that there will always be room for both formats in the industry.

AP: Paten began by stating that if foundational models didn’t improve at all from what exists today, the tooling is “really, really good”. Knowing that, and with the obvious understanding that FM’s are improving, he echoed Raza’s point about the opportunity in the application layer. While the winners in the foundational space will funnel out, companies building on top will compete with even the largest incumbents. Paten offered Perplexity, the AI-powered search engine, as a great example. The search experience, built atop OpenAI’s FM, “outmaneuvers” Google’s own search, and OpenAI’s product suite.

VM: As for how “mass adoption” will shake out? Mugunthan understands that the value will come from access to data. The accompanying challenge? Ensuring security, compliance and privacy when building solutions around that data.

LC: As an investor, Chang is particularly excited to support companies who are building the infrastructure for these applications. She painted the picture of how AI applications can help eliminate biases in systemic processes, if done right.

To close the panel, Ramakrishnan asked each panelist what they would be building or investing in, if not in GenAI. While it’s nearly impossible to picture a non=GenAI pick, they shared their answers below:

HR: “Probably space tech”.

AP: “A GenAI application for investment managers”.

VM: “Fashion tech or sports tech.”

LC: “Any company that could feed me on demand”.