John-Isaac Clark, CEO at Arturo discusses the application of Artificial Intelligence and Machine Learning for effective property analysis and geospatial technologies.
1. Tell us how you came to be the CEO at Arturo. How much of your typical day involves innovating or ideating Arturo products for your customers?
I was recruited by the American Family Insurance team to spin out the technology that is the basis for our product today, which they had made a substantial investment in over a two-year period. I’ve been a high-tech entrepreneur and worked with many startups in my career, including T-Sciences, which exited to Hexagon Geospatial. Through that business, I had the opportunity to work with Google in 2007 shortly after they acquired a company called Keyhole, which became the product we all know as Google Earth. I was fascinated by how consumers have so much incredible access to location-based information (much of it derived from images of places with AI), but I always believed there was a huge market for this same approach to be applied to enterprise business use cases.
When I saw how American Family Insurance had built technology that employed AI to extract information about properties from satellite images, aerial images, and ground-level images, I was very excited about how this could transform the way in which industries that insure, lend, invest, or manage properties make decisions and conduct their business.
During the first year, I was heavily focused day to day on our product development and ideation. In my role previous to Arturo, I was Head of Commercial Product at DigitalGlobe (now MAXAR), which is the world’s highest resolution satellite imagery provider. I have had the privilege of spending a lot of time listening to what customers say they want and ultimately working to understand the problem they are trying to solve or the outcomes they are trying to drive. I still interact daily with our customers, though I am fortunate to now have some incredibly talented people who focus on our product and customers here at Arturo as I work to continue to grow the business.
2. What are the applications and opportunities Arturo offers insurance providers with its deep learning and predictive capabilities?
We are enabling a transformation in the way businesses understand physical property by applying AI and machine learning to remotely sensed data collected on and around properties. In layperson terms, this means extracting valuable information about a property from images of the property taken from satellites or airplanes. In the future, we believe that our capabilities will drive similar value in adjacent verticals (such as lending), but we remain laser focused today on property and casualty insurance because we continue to learn so much from our insurance customers that can help us create value for them today, and for other industries tomorrow.
Fundamentally, we are driving substantial value for our customers in a few discrete areas—though when employed in combination, these returns can become exponential.
Since we are able to fetch available images of a property and return an analysis in seconds, we are often initially used in time-sensitive transactions where a human is waiting for the result. This could be a consumer requesting a new quote for a homeowners insurance policy, an underwriter validating the accuracy of a property’s policy information that came from an agent, or a claims adjustor wanting to both quickly assess damage and understand the condition of the property or structures prior to the loss event. In all these use cases, we are accessing current (generally within the last 90 days) high-resolution imagery (from multiple sources and providers) of the property and in around 7 seconds returning up to 65+ discrete characteristics regarding the property.
For the consumer, we may identify that you have a pool, a trampoline, and solar panels and assess the current condition of your roof in order to get you the best price. For the underwriter, we’re noting whether the pool is fenced and if trees and foliage are well maintained and away from the roof or building (in wildfire-prone regions) to better assess the risk factors present. For the claims adjustor, we’re determining the actual areas on the property or roof where damage has occurred and providing measurements of the damaged areas in order to quickly estimate and pay out the claim.
Our second way of working with insurers is via large-scale analysis. Instead of returning one property in 7 seconds, we can analyze millions of properties in very short periods. We recently made an announcement regarding our work with Suncorp in Australia, where we analyzed approximately 8.7 million properties (nearly every residential structure in the country) in 48 hours. That’s a rate of about 50 properties per second!
The many use cases here are different (but potentially have a far higher ROI), and include finding all properties that have poor conditions (risk identification) or those that have excellent conditions (marketing), identifying pockets of hidden risk within a portfolio (portfolio risk management) to diversify the risk through reinsurance, analyzing every property post-catastrophe within a large area to pre-determine the amount of damage or loss and quickly aid the homeowner (claims management and automation), creating a more efficient pricing model (pricing), and analyzing each property that has a policy that is up for renewal to make sure a property that has a major issue or risk isn’t automatically reinsured without remediation (renewals).
3. How did you define the vision of Arturo? How did you approach your first 100 days as the CEO at Arturo?
I think the vision really goes back to some of my early comments regarding my personal passion to see AI and machine learning on remotely sensed data (often images) of property that commonly improves our lives as consumers and improves outcomes and returns for enterprise businesses. I’ve also learned throughout my career that it can be a real struggle to do more than one thing really well.
One of the mantras I consistently repeat here at Arturo is that “we want to best at what we can be better than anyone else at.” That statement has been a guiding principle when I think about both the vision for the company and how we continue to pursue our goal. What set of activities can meet these criteria? What does Arturo intrinsically have that others do not that enables us to achieve this? If we are contemplating a new undertaking or customer ask, does it fall within these parameters or is it a commodity that is easily replicable or (worse) easily done better by someone else?
A vision is not a strategy, a goal, or a product. A vision is the impact your organization has on the world and the people that live (or will live) in it.At Arturo, our vision is to create a greater understanding of physical spaces in order to benefit and enrich the lives of those who own, occupy, use, or travel through them. Today, this vision is being accomplished by focusing very much on residential homes and the properties they sit on.
What are the features of the property? What is the condition of the house? What inherent risks do the physical features and conditions of objects on the property present? How could those risks be mitigated? What is changing on the property? Are conditions improving or declining? What is happening on the broader surrounding properties (neighborhood)?
By answering these questions effectively and accurately, and entirely without humans, we enable insurance companies to better price insurance, respond to claims faster, and ultimately mitigate risk, which improves the lives of the homeowner and occupants of the property. While this is how we are accomplishing our vision today, you can extrapolate this approach to other verticals that are real-estate related and other types of property (commercial properties such as retail, office, warehousing, light industrial, healthcare, schools, etc.) and achieve similar value creation.
As a spinout from a Fortune 500 Company (American Family Insurance), my first 100 days at Arturo were very much focused on trying to create a common sense of identity distinctly our own.
4. What are some distinctive key features of Arturo’s on-demand property analytics tool, and how do you differentiate yourself from your competitors?
We have three key differentiators. First is our proprietary access to insurer data, starting with American Family. This gave us highly unique training data as well as real-world loss data to empirically test and measure the performance of our models. This produced “data gravity,” which enabled us to obtain additional data from other customers as we have continued to grow, creating a flywheel effect that ensures we have the most performant models that are consistently measured against multiple customers’ data.
Second, we’ve focused on using multiple sources of imagery versus a single provider or source. This has required an investment in high-model transference, but our customers consistently tell us that they want to be able to leverage as many new image sources as possible. Today, this means we can use aerial imagery from more than four providers; tomorrow, it will allow us to incorporate drone imagery provided by customers (or a third party), imagery from stratospheric balloon imaging platforms (and we’ve already announced key partnership deals with Near Space Labs and UrbanSky, two leading providers in this area), and imagery from far higher satellite imaging platforms.
Last, we’ve worked hard to develop a feedback loop process that we’ve dubbed Full-Loop Deep Learning in order to consistently measure and improve our model performance via customer and consumer interactions. While this should be table stakes in any machine learning business, the reality is that this is very infrequently employed in offerings today. We want our customers to understand that this self-measurement and improvement is central to how we approach machine learning and is automated and integrated into the core of our approach.
5. Does Arturo plan to expand beyond the insurance domain to other industry sectors that may need AI-driven spatial data?
While we remain laser focused on residential Property and Casualty (P&C), we have substantial investments around commercial P&C, as well as residential lending, and are excited to make some announcements around these areas in 2021.
6. Arturo was recently in the news for joining hands with the Open Geospatial Consortium (OGC) community. Can you elaborate on what Arturo’s contribution will be in this collaboration to solve global challenges?
While we deliver structured data (think CSV or XML) to our customers, what occurs is an incredible amount of computer science. Much of this is rooted in geospatial technologies since we are answering questions about physical property in the real world. Our OGC membership represents our desire to give back to this community and to stay abreast of, and contribute to, emerging standards and methods to continue to drive innovation and advance this field.
7. What are some of the common pain points that your customers approach you with?
Their desire to automate their underwriting processes, eliminate (or optimize) physical property inspections, focus their people on the hardest problems that human brains are still the best at addressing, automate out repetitive processes, obtain more accurate information faster and cheaper, increase customer satisfaction, and remove friction (from quote and claims processes)—these are all common things we help our customers achieve.
8. Arturo has partnered with Near Space Labs and LexisNexis (LN) to build more comprehensive offerings for their customers. Are there any more partnerships in the offing?
I’ve commented on the relationships with Near Space Labs and UrbanSky, so I will focus on our LN announcement. As I mentioned, at Arturo we want to be best at what we can be better than anyone else at. Since we have heard customers repeatedly ask about synthesizing weather data, claims data, and highly accurate roof condition assessments, we saw an opportunity to partner with LexisNexis (which has industry-leading solutions with weather and claims data). This partnership allows us to abide by our guiding principles while meeting a market need. We will continue to pursue similar strategies with other category leaders as we scale our growth in insurance and other verticals.
9. Is a tech partnership a better approach than building out every capability within the Arturo software?
Absolutely. Again, we want to best at what we can be better than anyone else at. This means looking very hard at buy versus build and only building those things that we can do better than anyone else.
10. What advice would you like to give to the upcoming AI-based tech startups?
Focus first on your customer, the problem they have, and how to really solve it.Try to measure from day one the real business value you are creating for your customer. Relentlessly focus on making sure you consistently increase that value. Listen, listen, listen to your customers! Arturo would not be where we are today were it not for listening to incredible feedback and direction from our friends at Hippo Insurance, Openly, Branch, AmFam, and others.
11. Can you give us a sneak peek into some of the upcoming product upgrades that your customers can look forward to?
Temporal property analysis (being able to assess a property over a period of time) within our large-scale analysis capability is something we’re really excited about but haven’t publicly announced. We look forward to releasing this in our API as well so customers can perform on-demand analysis to assess individual properties in early 2021. We will be making some additional exciting announcements once we exit private beta with a few key early adopters whose feedback and partnership we’re really grateful for.
12. Which is the one AI breakthrough you will be on the lookout for in the upcoming year?
One-shot learning would be amazing but is likely farther away than we would all like.
13. What is the one leadership motto you live by?
Listen more than you speak, always be open to learning you are wrong, and your people and your customers are the most valuable thing you have—always put them first.
CEO at Arturo
John-Isaac Clark, CEO at Arturo: Former Head of Commercial Product at DigitalGlobe. 10+ years startup background in geospatial and location-based analytics. Recovering software engineer.