The DTC Interview: Distyl AI’s Arjun Prakash and Derek Ho

Working behind the closed doors at Palantir, Arjun Prakash, and Derek Ho saw firsthand how AI could transform an enterprise business. Yet, they watched across the Valley as one AI tooling startup after the other suffered a similar graveyard fate.

“Then, at the same time, serendipitously large language models became a thing, OpenAI truly and fundamentally changed the game,” said Arjun Prakash, CEO of Distyl AI. “DALL-E and ChatGPT created a delight factor and unprecedented user demand. Now people are asking for this transformation.”

Having spent years working alongside each other, Prakash and Ho always knew they wanted to create a business together. With a groundswell in interest and now enterprise-ready AI models, Prakash and Ho co-founded Distyl AI to help enterprises integrate AI into core business workflows and find value in unstructured data. Before announcing a Seed funding co-led by DTC and Coatue, we sat down with the co-founders to understand the technological inflection point and motivation behind their endeavor.

WHY THIS, WHY NOW?

Q: AI has been buzzy for a while, coming and going. What changed recently?

Arjun Prakash (AP): It’s interesting because the zeitgeist is largely discussing generative applications, B2C applications and some SaaS stuff. That’s not where we’re seeing the most value. The most value we’re seeing is in what we call information retrieval. But that’s also because we’re Silicon Valley nerds and have computer science terms for everything! Like most enterprises today have knowledge workers. And when you look at where the knowledge worker is spending most of their time, it’s not in making decisions, it’s in gathering the right information to make those decisions.

And we can take a few examples to be specific. If you were to ever call the customer service agent, for example, if you were a passenger at an airline. The conversation might be 10 to 15 minutes long. But most of the time, the agent puts you on hold to go through different systems to get the right information to help you. And the value that we see in using large language models is in providing a natural communication interface for this knowledge worker, in this case, the customer service agent, to get all the information from across the enterprise in a moment’s notice. So, they can put you on hold for two seconds, not ten minutes.

GPT3.5 came out, it introduced two new capabilities: one was question answering and the other one was information retrieval. This was game changing for large enterprises because at the end of the day, they have a lot of complex data sources and business processes. — Derek Ho

Derek Ho (DH): Six months ago, when GPT3.5 came out, it introduced two new capabilities: one was question answering and the other one was information retrieval. This was game changing for large enterprises because at the end of the day, they have a lot of complex data sources and business processes. They’re less interested in the generative aspect and more interested in how they can easily get information to make the right decisions. Arjun and I realized that this was the enterprise moment for AI and that’s why we decided to start this company.

(AP): I’ll add that everybody has talked about the value of AI for a long time, almost two decades now. But it’s historically only been accessible to people who understood AI. About a year ago, that changed for the first time with DALL-E when every single individual, whether you were technical or not, could get a delight factor from AI. With ChatGPT, again, everybody’s getting the delight factor from AI without having to understand what or how AI is built. What language models have done is they have now made AI accessible as a delight factor to anybody when historically, the only people who could get the delight factor of AI were people who knew how to build AI. And this sociological shift fundamentally makes it more primetime for mass consumption.

TALKING TECH

Q: What has changed that we’ve arrived at this point with AI?

(DH): If we can look at the longer horizon, like ten years ago, when people talk about AI, they’re talking about model architecture for building a model. For enterprises, this meant we needed to hire people to build these models. Then five years ago, people realized that the model is good enough, but we needed more training data. Then you can see a huge wave of startups trying to solve the bottleneck of data. If you fast forward to today, with the large language model, both the model and the data prompt have been solved. We are entering a new phase: how do we provide instruction to the model that brings in all the internal context of a specific company? This is creating a new motion and is why we are very excited because this is the moment for large enterprises.

LEARNING & INSPIRATION

Q: You both were at Palantir for a long time and worked with some startups—what inspired you to start a company?

(AP): It was very obvious to both Derek and I that we would do a startup together for the longest time. We’ve known each other for a long time and knew we are very complementary [of each other], in the best of ways, and that allows us to ensure that we have the right combinations for the company’s best interest. And so then the question is, why did we do this?

Honestly, it was out of necessity. We saw what a deeply integrated approach could accomplish in terms of transformation at a large enterprise in the way that Palantir undertook. After we left Palantir, we saw there was a missing element of integration and collaboration in AI. Everybody in AI was building tools available to enterprises, but enterprises didn’t necessarily know how to create the most value from those tools. That’s why there’s a graveyard of tools and very little transformative value outside of a few Silicon Valley companies.

…we saw there was a missing element of integration and collaboration in AI. Everybody in AI was building tools available to enterprises, but enterprises didn’t necessarily know how to create the most value from those tools. — Arjun Prakash

Then at the same time, serendipitously large language models became a thing, OpenAI truly and fundamentally changed the game. DALL-E and ChatGPT created a delight factor and unprecedented user demand. Now people are asking for this transformation, not just casually interested.

Q: What are your learnings from Palantir that you are pulling forward into Distyl?

(DH): I think the Valley loves to have a debate about Palantir. What kind of company is Palantir? But, unequivocally, some of the most important institutions have produced billions of dollars in value through a significant transformation that Palantir helped produce; the only way we could make that happen is through customer obsession. Number two is learning to sprint at the problem that is in the customer’s best interest. Execution was something that we did very well at Palantir.

And then the last thing that Palantir does really well is to create a culture of autonomy and leadership at every level. Every single person at Palantir is truly an owner of the company and an owner of a mission and an outcome that matters. If you hire 10x talent, you should give them ownership of 10x outcomes. Palantir did that very well, which we really would love to create because there are enough 10x outcomes to be created.

Q: Outside of technology—where do you find inspiration?

(AP): There’s this very famous video of Jeff Bezos. It’s a montage of him saying, “Customer first, customer first, customer first.” Literally a hundred different times over a 10- to 20-year timespan. And he always says the same line, “We built from the customer backward. We built from the customer backward.” We take a lot of inspiration from that because we’re very customer obsessed.

To me, the most inspiring people are people who listen. Good listeners. Where I grew up in India there’s this concept of a kirana store. A kirana store is a mom-and-pop store that accounts for 90% of the retail landscape in India. Every corner has a retail store, a kirana store. The closest equivalent in America is a bodega in New York City. They’re small shops, but they make their business by understanding the hundred people who come to them very well. They know if you’re going to want this biscuit next month. They’re already ordered it for you. We find that very inspiring.

We want to be in a position where we already know what our customer wants before the customer wants it. The bar we hold ourselves to and the people we find the most inspiring are people who listen to customers, understand what they want, and display a lot of empathy. — Arjun Prakash

We want to be in a position where we already know what our customer wants before the customer wants it. The bar we hold ourselves to and the people we find the most inspiring are people who listen to customers, understand what they want, and display a lot of empathy. And that’s not necessarily a business thing, but it’s just a good thing