The Network Gap
Your AI works and the company still fails. It’s a pattern common in “traditional” sectors — insurance, logistics, healthcare, energy. The hard part is not technical. It’s the long sales cycle, the regulation, their lack of urgency, and the fact that your network includes no one who can get the buyer to move quicker.
The misconception
In traditional sectors, technical superiority is not enough. They don’t buy tech from Silicon Valley tech bros. They buy from people they trust, through processes designed to minimize risk. Cold outreach doesn’t work. Introductions and in-person pitches may. If your network is purely tech, you won’t sell — even if the product wins on paper.
What we’re seeing
Vivek built Excarta with DeepMind expertise — their models were able to predict weather cheaper and nearly as good as incumbents. But when DeepMind and Nvidia open-sourced their weather models, the competitive advantage evaporated overnight.
The pivot revealed something: companies didn’t care about normal weather events. They wanted weather black swans — rare, high-impact events that are hard for ML to predict.
Insurance (property and casualty) got interested in hailstorm predictions. Not because they’re common, but because they’re costly and localized: 2-3 hailstorms per day in the US, over $50M in damages each, hard to predict with standard models. Trial users appreciated the functionality — “it saved my car from hail”. Insurance companies were prepared to sign high-value contracts.
But the contract legal review stalled and the company had burned its runway. Sales cycles run 12 to 18 months. Other clients wanted proof of an existing contract before signing their own — a chicken-and-egg problem.
The deeper problem was access. Vivek’s network was in tech. The decision-makers were somewhere else, behind gatekeepers — hard to reach.
Vivek noticed: “Some skills transferred very well from DeepMind Applied: structured thinking, planning, metrics, and defining how to measure success.” But he also notes: “Execution is where it’s difficult. Tech is easy.”
That’s not humility. It’s a diagnosis. In traditional sectors, the bottleneck is navigating industries where your credentials don’t open doors.
What to do instead
1. Hire for network before engineering. The technical founder has the technical stack covered. What they should focus on is optimizing the path to the buyer. Someone who knows how to get introductions to decision-makers. Someone who understands how those decisions are made. Someone who can borrow credibility you haven’t earned yet.
2. Fundraise with customers, not conviction. One path: raise first, figure out the problem later. The other: nail the problem, collect credit card numbers, then fundraise with leverage. Consider the second path. Investors can’t evaluate your access; customers can prove it.
3. Qualify the sales cycle before committing. A 12-month qualification process will kill you faster than a competitor. Before entering a traditional sector: Can you reach buyers directly? Do you have anyone on the team who’s closed deals in this industry before? Are you connected to the people who can push it forward when legal review stalls? If not, you’re not ready.
The diagnostic for leaders
Before entering a traditional sector with an AI product: Do you have a domain network, or just a tech network?
Big Tech talent plus product validation is not enough. It’s not even an advantage unless you can access decision-makers who don’t live on Twitter.
If you’re betting that technical superiority will overcome structural disadvantage, you’re not selling AI. You’re waiting to learn the lesson the expensive way.