Why we invested in TensorFlight: leveraging AI to create a more sustainable industry for all

By Alex Taylor, Global Head of Emerging Technology, QBE Ventures. 

Founded by ex-Google and ex-Microsoft engineers, Tensorflight is a leader in secure, ethical use of property data and artificial intelligence (AI). By leveraging its capabilities to help commercial property insurers make faster and superior risk decisions, Tensorflight is making insurance more sustainable fair and accessible for customers, insurers and communities.  

A common problem to insurance sustainability

What is often missed by industry insiders and outsiders alike is that the true size of the commercial insurance market is unknown. Aon sized the US market at $200 billion dollars when it purchased CoverWallet in late 2019. In comparison, Hiscox sized the market at $130 billion dollars in its recent interim results. There are several reasons for the discrepancy. However, our key hypothesis that a core, underlying problem exists regardless of the headline number: no one knows the true size. This is largely because insurers can’t get the ground-truth data to understand this and the risks they cover. 

Insurers size their market by assessing ‘who fits into our risk appetite and product range?’. Commercial property insurance is no different. The quirk of the sector is that a physical site inspection is usually required to accurately price risk - a person must go to the site, often with clipboard in hand, and make an assessment. The problem is obvious. At a certain point, the unit economics don’t stack up. The risk might justify the inspection, but the premium may not.

The result: small businesses are priced out, and large businesses face unnecessary delay and cost. Insurers balance uncertainty. Imagine the scenario below.

Scenario: a client with 50 worksites or a retirement village of 100 dwellings approaches an insurer. What option should the insurer select: 

  • Option A: Visit each site to provide underwriting insight for the policy
  • Option B: Visit a few sites and project insights forward onto the remainder
  • Option C: Visit all properties virtually and extract features, that can then be verified with the customer

Neither A or B are sensible, nor do they provide the best possible outcome for the customer. Fundamentally, risk is priced on uncertainty – and both these options provide substantial uncertainty. Far better if all properties can be visited (virtually), a feature extraction performed – to allow for sensible underwriting decisions without breaking the bank.

How TensorFlight is changing the game

Faster, cheaper, and better. TensorFlight’s mission is to automate the slow and expensive process of in-person property inspections. Founded by Dr. Zbigniew Wojna, a recognised world leader in visual feature extraction and machine learning, and Robert Kozikowski (ex-Google and Facebook), Tensorflight stands out in a class of its own. 

In a future of increased weather volatility, we look forward to working with TensorFlight to better understand and meet the insurance needs of customers. But the real ‘ah ha’ moment and realisation of broader potential hit when analysing post-hurricane disaster images. While most buildings were razed, one was virtually untouched. With no eave, the building could sustain the pressure from severe winds. This simple yet powerful insight illustrates how when used ethically, computer vision empowers insurers, customers, and the property ecosystem to better design, assess and protect properties at scale.

It also highlights what makes Tensorflight unique - while many companies can analyse satellite imagery, few have managed to make inroads into the automated analysis of oblique imagery, a fundamentally much more challenging proposition. Feature extraction from oblique (side on, arbitrary-angled ground-level) imagery has much greater promise. From an this vantage point, wall composition and condition, points of ingress and egress, and window condition and quality can be determined. Through this technology we’re on the cusp of being able to automatically determine what a human assessor can perform looking at the same imagery – but at much greater scale, and (with the right image source) without needing to visit the property at all. 

We invested in Tensorflight because they are leaders in the secure, ethical and sensible use of big data and artificial intelligence. Together, we’re excited to be powering forward-looking insurers to be there for their community and customers, before and after natural disasters, events, and accidents. 

Learn more about Tensorflight