Insurance Companies Aren’t Spying on You. They’re Trying to Send You a Warm Blanket.

Marty Smuin’s career reads like a tour of every major inflection point in commercial technology, in order. He was there for the birth of e-commerce at QVC, moved with Barry Diller into the IAC world building out Ticketmaster, CitySearch, Expedia, Hotels.com, and Match.com, then pivoted into wireless — solving, in the early days, the genuinely hard problem of getting video to compile in bits and packets over a phone network, working with NTT DoCoMo, AT&T, Verizon, and Singtel. He later ran one of the world’s most advanced high-performance computing companies, helped a friend build counter-drone radar detection technology for stadium security, and built a company serving small medical practices before deciding to step back and just serve on boards.

That plan lasted until he joined Arturo’s board, got pulled in by both the technology and the industry, and ended up leading the company. His framing of what drew him in: insurance sits inside a much larger real property ecosystem — county tax assessors, mortgage companies, land developers, contractors — all of whom benefit from the same underlying technology, at a moment when carriers are finally starting to move off platforms some have run for 35 years.

In Episode 123 of InsurTechTalk, Marty and I covered how Arturo actually extracts property risk data from aerial imagery, why he thinks the “insurers are spying on you” narrative gets it backwards, and where the MGA relationship fits into how large carriers actually innovate.

About Marty Smuin

Marty Smuin is the CEO of Arturo, a roughly five-year-old company providing AI and machine learning-driven property insights and analytics to insurance carriers, primarily in the US and Australia. Arturo doesn’t collect its own geospatial imagery — it draws from satellite, aircraft, drone, and balloon sources (the majority from aircraft-based aerial surveys) and applies proprietary AI models to extract structured risk data. Before Arturo, Marty built his career across e-commerce, wireless technology, high-performance computing, aerospace security, and healthcare technology, without a background in insurance specifically.

What Arturo Actually Extracts

The scale of what the platform pulls from a single image is the headline number, but the underlying discipline is what makes it useful for underwriting.

• Arturo currently extracts roughly 160 to 170 distinct attributes from a single piece of geospatial imagery
• Examples: roof material, rusting, ponding, slope, and overall condition score; number of chimneys; presence of a pool, hot tub, or sports court; AC unit condition
• Every one of these feeds directly into underwriting risk assessment — a damaged roof facing an incoming storm carries a materially higher claim likelihood than a well-maintained one, and pricing that difference accurately upfront affects the entire downstream cost structure of the policy

Marty’s summary of why the accuracy bar matters as much as the attribute count: Arturo runs roughly 35 to 40 different AI models, achieving 90-95% accuracy per model — a number that sounds low next to a telecom-grade “five nines” reliability target until you consider the actual scale. Arturo isn’t scoring a thousand properties or even a hundred thousand; it’s processing millions of images, and applying that accuracy consistently across millions of properties produces a genuinely reliable aggregate signal even where any single assessment might be imperfect.

Differentiation in a Crowded Geospatial Category

I pushed Marty on how Arturo stands out in a category with real competition — the geospatial vendors occupy an entire dedicated row at ITC Vegas, and the space has already seen its share of acquisitions and failures.

His answer came down to two things beyond the core models themselves:

• The underlying platform — Arturo builds on Esri’s GIS tooling rather than building geospatial infrastructure from scratch, layering its proprietary risk models on top
• Data enrichment beyond imagery — weather data, seismic activity, live brush fire tracking, nearest fire hydrant location. The platform is highly visual: a heat map of a carrier’s entire book across the US, with drill-down to 5-7 centimeter resolution on an individual property, and predictive overlays — for instance, projected storm surge height against the specific policies that fall within it

That predictive, proactive layer — not just describing a property’s current condition but projecting what an incoming hazard event will do to a book of business — is what Marty credits for Arturo’s recent win rate against competitors.

”Meet Your House”: Turning the Underwriting View Back on the Homeowner

This was the most interesting part of the conversation, and a genuine point of differentiation in how Arturo positions its product on its own website — speaking directly to the eventual policyholder rather than only the underwriter or claims adjuster.

The observation driving it: an insurer’s imagery might show 40% tree overhang on a roof, debris in the gutters, an unfenced pool, or trees growing dangerously close to a structure creating fire exposure — and the insurer sees all of it immediately. The homeowner typically has no idea any of this is being assessed, or that it’s costing them.

Marty’s pitch is to close that gap: let the carrier show the homeowner what the imagery actually reveals, and what fixing it is worth.

• Trim overhanging branches, or don’t get the discount
• Create a defensible fire perimeter around the structure
• Fence and fill the pool
• Replace visibly damaged shingles

His framing directly rebuts a Wall Street Journal narrative he referenced about insurers “spying” on policyholders via geospatial imagery: there’s nothing covert happening, and treating the underwriting view as a private secret only benefits the insurer. Sharing it turns a one-directional assessment into something both sides win from — better rates for the homeowner, fewer claims for the carrier — provided the industry is willing to open that channel rather than keep it opaque.

Who Pays, and Why the Homeowner Interaction Still Matters

Arturo’s paying customer is unambiguously the insurance carrier, not the homeowner — Marty’s rule of thumb: whoever pays for the service is the customer. But the homeowner-facing “meet your house” interaction still matters commercially, because monitoring, maintenance, and repair — catching a problem before it becomes a claim — is a genuine win for both sides of that relationship, not just a marketing gesture.

Why Carriers Are Moving Faster Now

Marty’s read on adoption timing is straightforward: catastrophe losses are forcing the issue. Hurricanes, tornadoes, floods, and fires are compounding faster than carrier balance sheets can comfortably absorb, and that pressure is translating directly into automation investment — underwriting, claims processing, fraud detection, and customer service all being pushed toward faster technology adoption than in prior cycles, in North America specifically.

We discussed the familiar tension this creates internally: technology is a cost center with a return that often doesn’t materialize for two or three years, which makes it a hard sell against the pressure for immediate bottom-line relief. Marty’s counter-framing borrowed from a story about his QVC days — selling $8 million of HP printers on air by demonstrating photo printing rather than describing it. His point: whenever you can remove the imagination required to understand a benefit — show it rather than explain it — adoption gets dramatically easier, which is part of why Arturo invests so heavily in the visual presentation of its platform.

Tier-One Carriers vs. Regional Carriers, and Where MGAs Fit

Larger carriers and smaller regional ones approach this technology from different constraints:

• Large carriers managing 5-10 million customers focus innovation spend on customer retention economics — reducing churn by even 2% materially changes the new-business target needed to hit growth numbers — and increasingly on customer experience: apps, portals, and clearer communication, partly driven by regulatory attention on issues like abrupt non-renewals after a claim payout
• MGAs have become an increasingly important channel for exactly this kind of innovation. As larger carriers exit certain geographies or risk categories, MGAs fill the gap, reselling products built by the larger carriers while moving faster and more flexibly on customer acquisition, engagement, and retention technology than the larger carrier could on its own
• Marty’s view on M&A as an innovation strategy: acquisition is one of the fastest ways to ingest new technology, but also one of the most dangerous, because integration is genuinely hard — a point I pushed back on directly from my own investment banking and integration experience. Technology integration between two companies, insurance or otherwise, routinely fails to reach its potential without the right leadership in place; the common failure pattern is acquiring a technology and either shutting it down or running it as a bolted-on external layer rather than genuinely integrating it

Closing Advice: Use the Wood Already Chopped

Asked for advice to someone entering the industry, Marty’s answer centered on not reinventing infrastructure that already exists:

• Change is coming regardless of whether any individual company embraces it — no one gets to be the sole agent of it
• Every decision should be backed by real data — proprietary data combined with third-party sources, not intuition
• Embrace machine learning and AI, but build on top of existing generative AI platforms rather than attempting to build foundational infrastructure in-house — that undertaking, at the scale of a Google, Microsoft, or AWS, is not a reasonable ambition for an insurance carrier
• Keep the actual goal in view: not the specific piece of technology or the company you might acquire, but customer trust and peace of mind — insurance, at its core, is meant to provide a “warm blanket” when something goes wrong, and technology should serve that, not replace it as the point of the business

Key Takeaways

• Arturo extracts roughly 160-170 property attributes per image at 90-95% model accuracy, and achieves reliable results at scale by applying that accuracy consistently across millions of properties rather than requiring near-perfect accuracy on any single assessment
• Predictive, proactive risk visualization — projecting what an incoming storm will do to a specific book of business — differentiates more than raw data extraction alone in a crowded geospatial category
• Sharing the underwriting view of a home directly with the homeowner reframes “insurers are spying on you” into a genuine mutual-benefit interaction: better rates for maintaining the property, fewer claims for the carrier
• Catastrophe loss pressure, not competitive pressure alone, is what’s currently forcing carriers to accelerate technology adoption across underwriting, claims, and fraud detection
• MGAs increasingly serve as the innovation and speed layer for large carriers retreating from specific markets or risk categories, reselling their products while moving faster on customer-facing technology
• Technology M&A is a fast way to acquire capability and a common way to destroy it — integration failure, not acquisition itself, is the real risk
• Building on existing generative AI platforms rather than proprietary infrastructure lets insurers focus innovation spend on the application layer, where the actual competitive advantage lives