Startup of the Day: Polish-American insurtech startup Tensorflight
Established in 2016, Tensorflight develops a property platform that reduces risk, waste, and cost based on convolutional neural networks. The technology connects to software via API and uses ground-level imagery combined with satellite and aerial imagery to create accurate replacement costs and better assess risk. Tensorflight claims that its platform allows to reduce 25% of costs for in-person property inspections.
Tensorflight’s team consists of over 30 employees and has offices in Warsaw, New York, and London. To date, the startup partners with some of the largest commercial property insurers in the world, including Zurich Insurance Group and QBE. In 2022 the startup closed $4.4 million in a Series A round led by QBE and FF.VC.
In the Startup of the Day column co-founders of Tensorflight Robert Kozikowski and Jakub Dryjas share more details about the startup’s idea, its product, and future plans.
“The Startup of the Day column on AIN.Capital is dedicated to tech projects from all sectors that originated from the CEE countries. If you would like to introduce your project, please fill in the questionnaire.”
Tell us about your startup. How does it work?
Tensorflight provides property insurance companies with high-quality data for more informed, accurate, and timely decisions. Our technology combines AI computer vision technology with satellite, aerial, and ground-level imagery to streamline and automate the time-consuming and costly processes involved with in-person property inspections. Additionally, we help insurance companies assess risk as part of the underwriting process and understand their portfolio exposures.
We are a Polish-American company with locations in Warsaw, New York, and London, and we are honored to work with some of the largest commercial property insurers in the world, including Zurich Insurance Group and QBE.
How did you come up with the startup’s idea? What was the reason/motivation behind it?
We saw the potential to create a technology solution that would address these needs and help insurers upgrade many of their slow, outdated processes involving property data.
Insurance is still one of the most traditional sectors, operating heavily on legacy systems. Particularly in the property insurance industry, there are many time-consuming, manual tasks involved in underwriting properties. There is also a general lack of accurate property infrastructure data.
How long did it take to reach the prototype or MVP? What did you encounter?
In early 2017, our company introduced “Orchard Intelligence,” an app developed for DroneDeploy, which was launched alongside their app market. The app was developed with a specific customer in mind, one that’s still with us until this day.
Approximately one year after our initial launch, we received our first outside funding. In 2017, we received $125.000 from Boost.VC accelerator program and we closed our $500.000 seed funding round. From this point forward, we decided to take a more holistic approach to helping insurance companies by collecting property data through three valuable sources: satellite, aerial, and ground-level imagery. This would give our clients access to more detailed, highly accurate information on any property, including data on the building construction type, roof pitch, and geometry, number of stories, detection of ACP panels, and much more.
Our product has evolved over the years – we had several MVPs over the years, but the MVP for our current product really materialized in 2018 and 2019. In 2019, we closed our first six figures contract with Nephila and Markel using our technology for the analysis of properties.
When exactly did you launch your product? Or when the launch is planned?
Tensorflight was founded in 2016, and in 2017 we launched our property inspection platform, based on convolutional neural networks, that connects to an insurer’s existing software via API. In 2019, Nephila Capital was the first company in the insurance industry to invest in our solution and since then, we have continued to release new products and solutions for underwriting, loss control and claims, and more.
Tell us about the startup’s business model. How do you monetize your product?
We monetize per “analyzed address.” By providing data about insured assets for insurers, we are helping them to better price risk. For example, by distinguishing between construction types, we can help insurers predict which buildings are going to get demolished in an event such as a hurricane. As a result, it incentives companies to construct better buildings. We are also pricing our services via API call per address. With each address, we can quantify insurers’ ROI in terms of avoiding bad properties in the future.
What are your target markets and consumers?
We provide high-quality data and analysis tools for insurers of residential and commercial properties around the world. More specifically, Tensorflight can be used by:
- Underwriters to access the data they need to perform more accurate and efficient analysis on properties
- Brokers to create differentiated offerings for their clients and insurers
- Claims agents automate many of the manual tasks involved in claims processing & settlements.
If the startup has already launched the product, what are the results: metrics, income, or any clear indicators that can be evaluated?
Today we are able to analyze 99% of properties in developed countries through our satellite, AI, and computer vision technology, replicating the work of human assessors on a much larger scale, without the need for physical inspections. This allows for a significant reduction in the cost and time spent on inspections and an enhancement of the claims process. On average, our solutions can decrease the cost of in-person property inspections by over 25%.
We more than doubled our revenue starting from 2021 to 2022, and early in 2023, we are already confident we will more than double it. We are working with 3 out of 5 largest commercial property insurance companies on multi-year, six or seven figures deals. In January 2023 alone, we’ve already closed three deals for several million in yearly revenue.
What about your team? How many people are working in the startup? If you’re looking for new employees, indicate whom exactly.
Our growing global team consists of over 30 experts from various fields, including data science, insurance, civil engineering, structural engineering, and architecture. We have evolved into a hybrid model, with our business and sales operations based in the US and our software engineering and research divisions based in Poland.
Currently, we have a few open positions on our IT team, and we’re looking for ML and DevOps engineers.
Have you already raised any investments? Provide us with more details on each funding round: the amount, investors, and the purpose of the investment.
In 2019, we secured a $2.2 million round of funding, as well as a $1.5 million equity-free government grant. In our latest round of funding, we raised $4.4 million (Series A) from QBE and FF.VC to help us grow our sales team in the US and Europe. We also patented our technology.
What’s next? Tell us about your future plans.
Our goal is to make property underwriting and risk assessment more efficient and cost-effective for insurers worldwide. Therefore, we are working hard on product development, scaling our team, and securing more key partnerships. Over the past year, we’ve partnered with leading digital risk processing platforms such as Cytora and Risk Solved. We also established a partnership with the Intelligence Business of Airbus Defence and Space, a global leader in Geospatial Data, Intelligence, and Defence Solutions, to ramp up our property analysis and assessment capabilities.
In 2023, we are expanding to the following areas:
- Massive scaling of our processing pipeline to tens of millions of properties
- Scaling of processing, ingestion, and standardization of broker standards like a schedule of values
- Incorporating the latest AI advancements into our product
- Expanding property attributes from 20 to 40+ AI-derived attributes.
You must be logged in to post a comment.