Building a startup in the AI era presents a paradox: early stage companies can scale rapidly — but they’re also more vulnerable than ever.
“It’s never been a more disruptive time. That’s the good news,” said Matt McIlwain, managing director at Madrona. “The bad news is, it’s never been a less durable time.”
McIlwain joined his former colleague Greg Gottesman, managing director at Pioneer Square Labs, on stage at Seattle Venture Day on Tuesday as the two longtime venture capitalists shared a clear message to founders: the AI era is rewriting the startup playbook, and sticking to old rules could leave even fast-growing companies behind.
New moats
In a world where large language models and open-source tools are widely accessible, defensibility looks different.
McIlwain described a concept he calls “reasoning flywheels” — a new type of moat where startups use data, engagement, and tailored workflows to build value.
The pattern works like this:
- Start with publicly available data and models.
- Build a superior user experience, grow customer engagement, collect proprietary data, and use that data to refine the models.
- That, in turn, improves workflows and generates even more data.
Gottesman noted that traditional defensibility based on ideas or basic IP is less reliable in this environment. “You have to think beyond where we typically thought about ways to build a moat,” he said. That includes exploring new pricing structures, delivery models, and user experiences — not just product features.
Unprecedented pace
The Seattle VCs drew comparisons to the dot-com era, when they began their investing careers. But AI, they said, is evolving at a breakneck speed that makes the internet revolution look slow.
“Even having done this for almost 30 years, I’m unfamiliar with this pace of change,” Gottesman said, adding: “The level of change and advance, and your ability to sort of envision what that future might look like, is unlike anything I’ve ever seen.”
McIlwain noted how models become obsolete rapidly, asking the crowd if anyone still uses GPT-2 — which was state-of-the-art not so long ago.
“The models, in and of themselves, depreciate — but intelligent applications and agentic workflows tend to have a flywheel effect,” he said.
McIlwain noted the critical importance of customer feedback in navigating such rapid change. “The customer is always smarter than the venture capitalists,” he said. “Listen to your customers.”
Gottesman encouraged founders to be “intensely, insanely curious” and stressed the value of a team. “Surround yourself with the most incredible people you can and try to fill holes for yourself,” he said. “Don’t just hire people that are exactly like yourself. Look for people that hopefully like you want to work with, but also that have different perspectives that bring in something new to your team.”
Bullish on Seattle
Both investors agreed that Seattle is well-positioned to lead the next wave of innovation, particularly in AI.
“I just feel like we have such incredible talent in the right areas — that infrastructure layer, as well as at the application layer around AI,” Gottesman said. “I think that puts us in a really good position to create one of the next great iconic companies.”
McIlwain said that if you believe the “reasoning revolution” will happen faster than previous technological revolutions — “how can we not, in the next 10 years, do incredible, transformative, daring things here in the greater Seattle ecosystem?”
Related: Can Seattle own the AI era? We asked 20 investors and founders to weigh the city’s startup potential