Let’s be honest: there’s very little actual technology innovation in InsureTech. Yes, we’ve seen genuine value being delivered, but with $30bn being invested by VCs in InsureTech since 2021, you would think that this would be an indicator of an industry pushing to create new frontiers. Instead, the industry seems content with taking ten-year old technology and repackaging this with a slick new user interface. Add in a decent budget for marketing, and the reality is a bit different to what those VCs are buying into. It works, and so this technology is still a benefit to the insurance industry. But we really should be calling it what it is. 

In defence of InsurTech, breakthroughs have been made. Just look at the likes of Lemonade – powering claims processing using AI has massively reduced the time it takes for a customer to get to where they need to be. Then we have Root Insurance – leveraging the technology in smartphones to influence driving behaviour, reward those for driving safely, and targeting a demographic that has grown up expecting this level of convenience in insurance. However, as an industry we need to temper our celebrations around “groundbreaking innovation” and recognise what these are: the geographic arbitrage of existing solutions, combined with superior user experience design.

Innovation or just shiny new packaging? 

To understand where real innovation sits versus incremental improvement, let’s define what constitutes genuine “tech innovation” in insurance.

Genuine Innovation – from where I sit –  involves developing new algorithms, novel data processing methods, or fundamentally different approaches to risk assessment. Examples include parametric insurance models that trigger automatic payouts based on weather data, or computer vision systems that can predict accidents before they happen. Repackaging, while valuable, takes proven technology and makes it accessible through better user interfaces, regulatory compliance, or market adaptation. 

Now, this can create real economic value. A great example being a mobile-first interface that makes insurance purchasing seamless, even if the underlying underwriting algorithms are unchanged. These changes are purely cosmetic. The problem, though, isn’t that repackaging exists. The problem is that we often label superficial incremental changes as more “revolutionary” than they actually are. 

The problem is stagnation.  As an industry we’re not raising the bar. We’re putting the bar in the wrong place and saying we’re already beating it.   

The stagnation of telematics

We see this pattern perfectly in telematics and AI dashcams. I’ve had plenty of conversations which back this up too – most recently a discussion with an insurance company selling a “revolutionary” insurance offering. What did it consist of? Continuous GPS data, accelerometer data, and location data – all available for over a decade in telematics systems. When pressed on how their offering differed from Progressive’s established $18 billion commercial lines solution in America, their response was telling: “Nothing. But you need to understand that Europeans, for whatever reason, haven’t done that.” A glittering example of innovation being mistaken for geographic arbitrage.

There is legitimate value in bringing proven American telematics to European markets – regulatory adaptation, local partnerships, and cultural customisation all matter – but we must recognise that this is not innovation. It’s straight from the business development playbook. 

Looking at core telematics technology, there’s been minimal fundamental innovation in a decade. We still collect crude GPS and accelerometer data from vehicles, correlate it with claims, and then run statistical analyses for predictions. Deep neural networks were introduced in 2015/16 to better establish correlation where statistical regressions missed the mark, but that really was the only evolution of this process.  

Ten years ago, that was a technological leap. In 2025, it’s old technology. 

The problem with dashcams

The rise of AI dashcams hasn’t fundamentally changed the equation either. These systems have never managed to produce underwriting-grade insurance. While traditional telematics data is part of the insurance process, AI dashcam data isn’t used for underwriting in any significant way in any part of the world. That’s quite revealing. These dashcams essentially add video event clips to tell you how many times someone was “harsh-braking” (data that comes from telematics anyway) and perhaps a few more clips from the cabin about distraction. And even with modern transformer architecture placed on top of this data, it may not make any real difference compared to the DNN architectures from 2015-2014.

Beyond the surface features, AI dashcam solutions rely almost entirely on off-the-shelf implementations. It takes mere seconds to implement capabilities like gaze estimation, emotion recognition, or face detection. These are widely available capabilities and, in many cases, it’s simply a single line of code which is being called to perform a function. It may look impressive when you see all these boxes and detection overlays, but the underlying technology is readily available to anyone. As a result, they all do much the same thing. 

An improved user experience

Playing Devil’s Advocate, these incremental improvements do improve user experience. Compared to a decade ago,  insurance is more accessible, understandable, and responsive to consumer needs. This creates genuine value, even when built on established technology foundations. Then there are the societal and economic benefits of geographic expansion of technologies. If anything, insurance has become more democratised – readily adapted for local regulations, cultural preferences, and market structures. This work, while not fundamental innovation, serves an important function in making the insurance experience largely better.

There are other success stories in InsureTech too. Embedded insurance – integrating coverage directly into purchase experiences – represents genuine innovation in distribution. Usage-based insurance, while built on existing telematics, has created fairer pricing models for consumers. Digital-first approaches have democratised insurance access and comparison shopping. These advances prove the industry can innovate meaningfully. The challenge is distinguishing  between solutions that fundamentally improve risk understanding, pricing accuracy, or customer outcomes versus those that simply offer more polished presentations of existing capabilities.

How do we understand true risk?

Consider this scenario: You’re driving and need to brake suddenly to avoid hitting a car rashly pulling out in front of you. Existing technologies such as telematics would not recognise the risk and possibly even deem your action bad driving. First generation AI Dashcams may classify the objects but not understand their intention. To move beyond this limitation, you need to read the road. And if you want to be doing commercial lines, you need to be able to read the cabin.

The industry faces several significant challenges. Firstly, data science and actuarial teams often speak across each other rather than to each other – with fundamental differences in how they measure and interpret risk, validate solutions, and measure success.

Additionally, creating systems that can actually understand contextual risk requires custom transformer models that are both powerful and efficient enough to run on edge devices, or in the cloud at price points that are not cost prohibitive. And it needs the team to implement and run them. There is currently a colossal shortage of technical talent in the insurance industry, and this talent gap is forcing less technical people to move into the technical. Or at least to repackage existing solutions with a more technical appearance.

The expertise gap in insurance stems from a fundamental chicken-and-egg problem. The industry’s natural conservatism on risk innovation make it difficult to build compelling business cases for hiring elite technical talent, who consequently flow to sectors offering greater opportunity for transformative work.

The path forward

True innovation in InsureTech requires moving beyond digital interfaces and off-the-shelf AI toward solutions that fundamentally transform how we understand and quantify risk. We need to put adequate resources – time and talent – behind developing custom technology that can analyse driving with full content. Only then, if we move beyond simple correlation, will we truly be able to predict (and prevent) outcomes. 

An InsureTech industry that analyses what is in front of the vehicle, and that breaks away from traditional  proxy measures like age, address, and occupation may be the only way to deliver meaningful value from new technologies while meeting stringent and varied regulatory demands.

The Automotive Risk Understanding and ADAS software markets are projected to grow from $21bn today to over $40bn by 2030. Capturing this value will require genuine technological innovation, not simply repackaging existing solutions with better marketing. As an industry, let’s aim for this level of excellence. Leveraging these technologies could deliver profound social good. We could empower consumers with information and support to reduce their risk, transforming the insurance relationship from merely pricing risk to actively preventing it. Road safety would improve as drivers receive real-time feedback and coaching rather than annual premium adjustments.

The insurance industry is ready for real innovation. It’s time we delivered it.