Log in

Forgot password?

Don't have an account with us?

Register Now
NEWMemberSpotlight-Feature (8)

Tensorflight: property data around the world

Tensorflight automates commercial and residential property inspections for underwriting, risk assessment and claims processing. InsTech’s Ali Smedley sat down with Tensorflight’s Maxi Wandel to discuss how the company’s solution can speed up the underwriting process and the partnerships it is looking for. The discussion also covered how Tensorflight is helping to assess building damage in Ukraine.

Member Spotlight: Tensorflight

What does Tensorflight offer to the insurance industry?

Tensorflight offers data and insights on commercial and residential properties for underwriting, risk assessment and claims processing. Using ground-level, satellite and aerial imagery, coupled with advanced computer vision models, Tensorflight extracts detailed insights about buildings and their surroundings. Property attributes offered include building footprint, square footage, number of stories, replacement cost, construction type, roof attributes and more.

Tensorflight’s data integrates into insurers’ existing workflows through multiple options including an API, a web application or batch processing for large property portfolios.

What geographies does Tensorflight focus on?

We provide our services to insurers worldwide. We initially started in North America and quickly expanded to Europe and Australia. Currently there are active projects in the UK, Switzerland, Poland and France. Our aim is to cover the entire European Union by the end of 2023.

How does Tensorflight’s solution help with the underwriting process?

With Tensorflight’s artificial intelligence (AI) technology, insurers can more accurately assess the properties they are insuring. This helps to ensure that the insurance coverage offered is appropriate and reflects the property’s value.

For small commercial properties, our automated solution provides an output within 4 seconds. Tensorflight analyses pictures of buildings from different angles and measures their features such as size, shape, material and condition, allowing insurers to efficiently evaluate the properties without the need for manual inspections.

In the case of large commercial properties, we add a manual verification process. Our specialist team verifies the data generated by our AI algorithms, resulting in an accuracy rate of 99%. Our platform is also capable of processing multiple locations at the same time, enabling the large property portfolios to be efficiently evaluated.

What data does Tensorflight offer around replacement costs?

Tensorflight provides comprehensive data on replacement costs for both commercial and residential properties. To determine this value, we use our AI and computer vision to analyse images of buildings and extract various features.

Our analysis encompasses essential factors such as the property’s footprint, construction type, facade material, number of storeys and quality of finishes. We regularly update our replacement cost data, taking into account factors such as market trends, inflation, material costs and labour costs. This helps to ensure that insurance coverage aligns with the current market conditions.

Our solution for main property features is available in all European countries, along with replacement cost data for key markets. We are continuously expanding our coverage and researching new markets. As Tensorflight expands, we are conducting in-depth market research to gather information on local factors in specific cities or regions. This approach allows us to provide more accurate replacement cost data tailored to each market.

What projects is Tensorflight currently working on in Europe?

We are currently engaged in several projects with large European and global insurers. We are providing insurer PZU with replacement cost data for small and mid-market commercial properties. Our collaboration aims to help PZU identify undervaluation of properties and ensure accurate coverage assessment.

We have started a project with Munich Re, focused on underwriting solutions for SMEs through its “flowin” platform. We are currently providing building attributes such as occupancy type, number of stories and COPE (Construction, Occupancy, Protection, and Exposure) data. Our collaboration began in France and we have plans to expand to other Western European countries in the coming year.

Tensorflight is also working with QBE, helping the insurer save weeks of work through identifying and filtering portfolio buildings in the UK with specific characteristics within 24 hours. Led by Simon Pink, the UK Head of Emerging Technology, we continue to collaborate with QBE underwriters and modellers to evaluate our ability to further complete, validate and enrich property schedule data.

In addition to these collaborations, we have been working with the Kyiv School of Economics to assess the damage caused to infrastructure in Ukrainian cities. This project is part of our commitment to using our technology for broader societal impact.

How does Tensorflight communicate the accuracy and confidence of its data?

We provide a confidence score for each attribute we return, which indicates the probability of the answer being correct. The accuracy of our data may vary based on the specific buildings analysed, as some insurers may have portfolios that diverge from our training data due to unique geographies or building types.

Typically, during a process for a prospective client, we examine a few thousand buildings. The outputs from our AI models are then compared to ground truth such as in-person inspections to ascertain accuracy. On average we achieve 95% accuracy for standard property attributes and over 85% for more challenging data points.

What new data is Tensorflight planning to offer over the next year?

Tensorflight is continually adding new property attributes. We recently introduced “estimated year built”, which uses historical imagery to determine the construction year when it may otherwise be unknown. We are also expanding our roof degradation analysis and our occupancy classification by including additional classes and tenant classifications.

Early-stage development is underway on “building survivability scores”, which will estimate a building’s resistance to different hurricane wind speeds. We are also enhancing our ability to manage complex multi-building parcels, helping to determine whether multiple buildings or parts of a building are insured.

We are currently testing the Tensorflight AI assistant with some clients.

What partnerships is Tensorflight looking to explore?

We are actively looking for partnerships within the climate and sustainability space. We want our expertise and data to help tackle the global challenge of climate change.

Tensorflight also partners with other technology companies and imagery providers. Throughout the last year, we have worked with organisations including Cytora, Socotra, Geosite, RiskSolved and Nearspace Labs. These partnerships involve integrating Tensorflight’s aerial and satellite imagery data into various platforms and solutions related to risk assessment and geospatial analysis.

We are also happy to announce our latest program, TensorConnect. It is aimed at resellers and makes Tensorflight’s property intelligence solutions more accessible. We will be publicly launching more information soon, but anyone already interested can contact us directly through our website.

What should readers do if they want to learn more?

If you want to learn more, please contact us at [email protected], or visit our website at tensorflight.com. Additionally, me and Daniel Buda will be attending ITC in Barcelona and would happily tell you more about Tensorflight’s capabilities.

Stay up to date with the latest from InsTech