The founder teaching insurers to read a house from the sky

The question did not sound like a business plan. It sounded like something you ask at 3 a.m. "If I will die tomorrow," Izik Lavy has said, "like, what is the value that I bring to the world?" Most people file that thought away and go back to sleep. Lavy built a company around it. That company, GeoX, now looks at 160 million buildings from above and tells the people who insure them what is actually there.

The path to that answer ran through an unusual classroom. Before GeoX, Lavy served in the Israel Defense Forces' Unit 9900, the intelligence outfit that specializes in reading the physical world from imagery - terrain, structures, the small visual details that separate a guess from a fact. He led an AI team there. The skill he came away with was narrow and rare: teaching machines to look at a picture taken from very high up and understand the ground truth underneath.

In 2018, at 22, he pointed that skill at a problem most founders find too dull to touch. Property risk. Roofs. Underwriting. He co-founded GeoX with Eli Lavi, who became chief technology officer, and Guy Attar. The pitch was not glamorous. It was enormous.

What a photograph from the sky actually knows

Insurance has a quiet, expensive habit. When a carrier writes a policy on a building, it often does not know much about the building. It knows an address. Maybe a year it was built. To learn more - the roof's age and material, its condition, its slope, whether the property sits in the path of a flood or a wildfire - someone historically had to drive out and look. That costs money and time, and it does not scale to a whole country.

GeoX replaced the drive with a photograph. Its software takes aerial imagery, reconstructs the property in 3D, and extracts the physical attributes an underwriter needs. The company reports coverage across 160 million U.S. commercial and residential properties, accuracy around 90 percent, and answers returned through an API in about a second.

Capture

High-resolution aerial imagery of the property

→

Reconstruct

Machine vision builds a 3D model of the building

→

Extract

Roof age, condition, slope and hazard exposure

→

Deliver

Confidence-scored data through the API

The way Lavy frames it, the product is not the imagery and not even the model. It is the insight handed to the customer. "Our mission is to help the insurance carriers to provide better service to their clients, to reduce the risk and also to provide a new revenue stream," he has said, "and the way to do it is by providing them insights, with data insights."

The contrarian bet inside an AI boom

There is a small heresy in the way Lavy talks about his field. Everyone around him is selling prediction - probabilistic models that estimate what a building might be, what a season might bring. Lavy keeps steering back to measurement. His argument, made plainly in interviews, is that deterministic data can outperform probabilistic AI when the stakes are real. When a hurricane is two days out, an underwriter does not want a probability distribution about your roof. They want to know what the roof is.

That conviction is easier to hold when the timing is on your side. The demand for GeoX did not appear from nowhere. Climate change has made natural disasters roughly five times more frequent over the past four decades, and the bill has followed. Insurance payouts for climate-related damage reached around $100 billion in a single year. Carriers that once treated property inspection as a cost center suddenly needed to understand every building in their book, quickly and cheaply.

Japan, and the millions of buildings

The clearest proof of the idea came from Japan. In 2022, GeoX began working with Sompo, one of the country's largest insurers, to scan millions of Japanese buildings and assess their exposure to natural disasters - a country where earthquakes are not a hypothetical. The alternative would have been armies of surveyors. GeoX offered a scan from above instead.

Lavy described the appeal in operational terms: "Our technology saves the need to send surveys to each house during the underwriting process at high cost, and to provide an accurate price assessment automatically to customers." The customer, in other words, gets a fast quote at a fair price, and the carrier avoids sending anyone to the door.

GeoX by the numbers

Properties

160M

Total raised

$23M

Series A

$19M

Rev. growth

~300%/yr

Accuracy

90%

Money follows the roofs

In September 2024, GeoX raised a $19 million Series A led by Flashpoint Venture Capital, with participation from Suretech and the investors Ariel Maislos and Noam Lanir. That brought total funding to about $23 million. The company said revenue had grown roughly 300 percent a year over the prior three years, driven largely by existing customers using more of the platform - the kind of expansion that tends to make investors comfortable.

The client list reads like a map of who worries about buildings for a living. Reinsurers such as Munich Re. Insurers like Sompo. The broker AON. The data firm Geoscape. And, on the public side, institutions including the World Bank and FEMA have made use of GeoX property intelligence to understand real estate risk. For a company that started with one person's late-night question, that is a wide reach.

The operator underneath the founder

Lavy is not only a technologist. Listen to him on the business, and a sharp commercial mind shows through. He talks about compressing the sales cycle to almost nothing - "You can make the entire sales cycle in one phone call" - and about the discipline of focusing on closing the next call rather than chasing the next lead. He returns often to a single idea he treats as foundational: "One of the biggest things that we need to do is understand our client better."

It is a tidy loop. A company built to help insurers understand buildings, run by a founder whose main rule is to understand customers. The product and the person are pointed the same direction - at knowing, precisely, what is actually there.

What he wants next is not modest. The aim is to make property risk a settled fact rather than a standing guess, and to make GeoX the default data layer beneath the global insurance industry - the thing carriers, banks and governments quietly rely on to know a building before anyone signs. Whether the market rewards facts over forecasts is still being decided. But the roofs are all still there, in plain sight, waiting to be read. Lavy figured out how to read them, and built a business on the reading.