The Story of DeepSet: Building the Foundation for AI’s Industrial Transformation
Six years before ChatGPT captured the world’s imagination, DeepSet’s founders saw a different future for artificial intelligence – one where AI would be embedded in every product and process, not just consumer-facing chatbots.
“Back then when we actually started, no one talked about llms, few talked about nap back then,” Malte Pietsch recalled in a recent episode of Category Visionaries. “And were in this early phase where it was with early signs that it comes easier to use these kind of transformer models.”
The Bootstrap Beginning Starting with just €5,000, DeepSet took an unconventional path. Instead of chasing venture capital, they built a profitable professional services business, working directly with enterprises implementing AI solutions. “We started basically doing professional services. We built custom AI solutions, learning solutions for enterprise customers,” Malte explains.
This early phase proved crucial. Working closely with companies like Airbus, they gained invaluable insights into enterprise AI needs. The profitability of services created a tempting path, but the team maintained their product vision: “We didn’t want to build a professional service company.”
The Open Source Evolution Two to three years into their journey, DeepSet launched Haystack, their open-source framework. This wasn’t just another product – it was a strategic move to build credibility with developers while maintaining focus on their ultimate commercial goals.
“We more or less from day one, then thought, okay, what’s the commercial product like? What’s the business model that can really create here to have a sustainable business?” Malte shares. They studied successful open-source companies like Elastic and Databricks, crafting a strategy that balanced community growth with commercial development.
The ChatGPT Acceleration The explosion of interest in AI following ChatGPT’s release created what Malte describes as a “washing machine” effect: “You wake up in the morning, you check, oh, what’s happened now? Again, where you have to very, I think, quickly adjust to where the market is going.”
But rather than chase the hype, DeepSet doubled down on production value. “Last year, I think that was where a lot of this noise happened, where so much new things became possible. So a lot of cool ideas, great ideas, exciting demos,” Malte reflects. “But I think at some point, if you don’t want to become hype and just a bubble, you need to show the value coming out of things.”
The Commercial Platform Today, DeepSet offers two core products: Haystack, their open-source framework for individual developers, and DeepSet Cloud, their commercial platform for enterprise teams. This dual approach creates what Malte calls a “sandwich motion” in enterprise sales: “It’s really a lot of developers out there know it, tried it, like it and typically like the devs are not our buyers… but they help a lot in this buying process.”
The Future Vision Looking ahead, Malte envisions AI becoming ubiquitous in enterprise software: “We are kind of in the middle of a huge industrial transformation where AI gets really embedded into almost every product, every process, and it’s really, I think, becoming this new generation of intelligent applications and processes.”
For DeepSet, success isn’t just about building technology – it’s about enabling this transformation: “In three to five years, I really believe AI will be embedded in all products, like in our daily life. All processes or products will have some AI feature embedded. And, yeah, hopefully a lot of them are powered by haystack, a lot of them are powered by cloud.”
This vision goes beyond mere product adoption. For Malte, success means creating a lasting impact: “One of the biggest motivation factors and boosts is if I get out, talk to random people, and they say, oh yeah, I know haystack, we use it for this and that use case, and we build something here and there for our product.”
DeepSet’s story suggests that sometimes the most transformative companies aren’t built by chasing trends, but by maintaining unwavering focus on a vision of the future – even when that future isn’t yet obvious to everyone else.