# 170 Jake Stauch, Co-Founder of Serval: Bet before the technology works, build infrastructure over raw models, and scale enterprise AI reliability

Jake Stauch is the Co-Founder and CEO of Serval, an AI-native platform that automates enterprise employee support through natural language-to-code workflow generation. Rising to prominence in the mid-2010s as a founder and product executive at the intersection of hardware and enterprise software, Stauch became known for identifying friction bottlenecks in IT automation and building infrastructure-first AI systems before the underlying technology fully matured. Serval, co-founded in April 2024 alongside CTO Alex McLeod, reached a billion-dollar valuation within 18 months of founding after raising $125 million across three rounds led by General Catalyst, Redpoint Ventures ($47M Series A), and Sequoia ($75M Series B).

Previously, as Director of Product at Verkada from 2019 to 2024, Stauch spent five years conducting customer discovery with enterprise IT departments across physical security hardware and software. There, he identified the automation paradox that would become Serval’s founding insight: despite a growing landscape of automation tools, most IT requests were still handled manually because the friction of building workflows exceeded the cost of doing the tasks by hand. His product work at Verkada spanned new product lines in physical security cameras, access control systems, and alarm hardware sold to Fortune 500 IT departments.

Earlier, Stauch founded NeuroPlus, a brain-sensing hardware and cognitive performance software company, which he led as CEO from 2012 to 2019. He was recognized on the Forbes 30 Under 30 list in 2017 for this work, which included a patent for an EEG-based neurofeedback system. He holds a degree from Duke University.

Listen to this episode on Spotify or Apple Podcasts

The infrastructure-first philosophy that helped Serval raise $125M at a billion-dollar valuation within 18 months of founding.

Jake Stauch left a comfortable Director of Product role at Verkada to start an AI company before AI coding was reliable.

“To be very clear,” he tells me, “the early version of our product did not work.”

I ask him about this and he does not flinch. He and his co-founder Alex McLeod quit their jobs and started building Serval when the best you could get from an AI coding tool was Copilot autocomplete. The vision was a system where you describe a workflow in natural language and the AI writes the code to make it happen. In the spring of 2024, that vision was aspirational at best. The models hallucinated. The outputs were unreliable. You could get something functional if you force-fed the system enough examples and kept the scope narrow, but production-grade it was not.

“It was so close to working,” Jake says, “that we had a lot of confidence.”

I have been listening to a lot of Founders Podcast lately, and one pattern David Senra keeps surfacing is the missionary founder -- someone who bets on themselves before the evidence justifies it. Steve Jobs spent $50 million bankrolling Pixar for 10 years with no business model, no market, nothing. Pure belief that the right team would figure it out. I share this with Jake and he grins.

The difference with Serval is that Jake was not betting blind. He had spent five years at Verkada watching enterprise IT departments struggle with the same problem: powerful automation tools that nobody used because building the automations took longer than doing the work by hand. He knew the market pain was real. The bet was on the technology catching up.

What fascinated me was how Jake thought about that bet. He draws a comparison to wireless communication, and it is the kind of analogy that changes how I think about infrastructure.

“Wireless communication is not very reliable at the physical level,” he says. “There’s a lot of loss of signal, loss of data, challenges and interference and all these problems. But we’ve built such robust systems that account for all of that and can mitigate all of that, that we have the experience that wireless is very reliable.”

He pauses. “I felt the same was true in the early days of AI. Even if it doesn’t get all that much better, I bet that we’re not even tapping into all the things we could do to build infrastructure on top of this to really take advantage of it.”

This reframes the entire AI startup calculus. Most founders I talk to are betting on the models getting better. Jake bet on building engineering systems that make existing models reliable enough to ship. The models improving was a bonus, not a requirement.

I bring up Spotify. Gustav Soderstrom, their Co-CEO and head of Product, talked about how Spotify’s early differentiator was not the streaming technology itself but a creative engineering trick: play the first 30 seconds of a song instantly from a smaller file, then use that buffer time to load the rest in the background. The macro trend of better internet connections would eventually make this unnecessary, but they did not wait for the trend. They built infrastructure to deliver the experience now.

Jake nods. “I think we’re reaching a certain frontier in a lot of ways, at least in the basic consumer interaction,” he says. “We haven’t even scratched the surface though on what you could do with the fundamental technology.”

He extends the wireless analogy further. The gap between basic radio communication and everything we do today with Bluetooth and WiFi is enormous -- and the fundamental physics have not changed. The innovation was all infrastructure.

The hardest part, Jake tells me, was deciding when to build for the models as they were versus when to wait for the next generation to make your work obsolete. In the early days -- summer and fall of 2024 -- every model update completely changed their assumptions. They leaned toward betting on improvement. Over time, that shifted. Now Serval builds to make existing models perform at the highest possible level, regardless of what comes next.

“You had to make all these decisions with imperfect information,” he says. “We generally leaned towards assuming that the models were gonna make most things better.”

There is a timing discipline here that most AI founders miss. The window between too early and too late is narrow, and it keeps moving. Jake caught it because he was not just watching the technology improve -- he was watching enterprise IT departments drown in the same problems year after year. The demand side was stable. The supply side was accelerating.

Eighteen months later, Serval has raised $125 million, reached a billion-dollar valuation, and is penetrating markets dominated by legacy players like ServiceNow. Not because the models got better -- though they did -- but because Jake and his team built the infrastructure layer that made unreliable technology reliable enough to ship.

“Man, over the past 20 years,” Jake says near the end of our conversation, “what all these software platforms could do outpaced anyone’s ability to actually implement and use them.” He says this almost casually, but it lands like a thesis statement for the entire AI infrastructure era. The models are powerful. The implementations are not. The companies that win will be the ones building the bridge.

Jake Stauch started building that bridge before the other side was visible. That is what missionary founders do.

Subscribe to the wayofproduct.com for more in depth guest profiles that are worth the time to read.

Get full access to The Way of Product w/ Caden Damiano at www.wayofproduct.com/subscribe