The insurtech aims to be the first to crack claims prediction using deep learning 

London based Insurtech has invested in deep learning in a bid to predict claims and on a wider level spot fraud.

It has also doubled the size of its team to around 15 staff with a particular focus on data science following a $2.5m funding round.

Niels Thoné, co-founder and chief executive at told Insurance Times: “Deep learning [as opposed to machine learning] has the challenge that it only works on really large data sets. It is very challenging to apply claims data sets to so you cannot get top performance.”

But the insurtech has managed to develop two proprietary pieces of software that enable deep learning that can be applied to smaller data sets.

It can automate 90% of claims, and Thoné believes that no other firm can do this in the same manner, the chief exec has high aims for to be the first to crack claims prediction.

“I believe there’s a secret race going on, and we want to be the first to crack it.”

Deep learning is subset of machine learning based on artificial neural networks that are capable of learning unsupervised from data that is unstructured or unlabelled, it is relatively new to the market. For example, machine learning would be something like Siri, but deep learning would be having a human conversation with Siri.

It follows the insurance industry speculating whether robotics such as Google Meena would aid the sector using deep learning.

The insurtech was founded in 2018 and aims to be a ‘plug and play’ option for the industry regardless of legacy systems to settle claims quickly after being inspired by the likes of Lemonade. claims it can settle claims in 24 hours, in the UK the average claims settlement is 25 days, although Lemonade once set a record settling a claim in just three seconds.

Bread and butter

After being incubated at London’s Imperial College and doing various tests, Thoné said: “We saw a lot of inefficiencies, leakage in claims, crude systems for short term processing.”

It has signed up eight customers so far and worked with accelerators TechStart and MetLife.

The insurtech began offering property and casualty insurance but has since expanded into health insurance.

When asked why it focussed on claims, Thoné said: “That’s really where the customer has its touch points with the company, and that’s really where the biggest issue lies.

”Claims is the ‘bread and butter’ of insurance companies. But it is also the most inefficiently run part of the organisation.

“As an entrepreneur I realised that this is a problem that affects the customer and the company. The problem is that you have a very limited data capture at the beginning of the journey – there is very little to make an accurate prediction.”

Firstly, the claims handler must fetch the information, check it, and then wait for other information from the loss adjuster. This, Thoné said, is holding up the process.

Using natural language processing can extract this and have access to a larger data set.

“Now you have a bigger data capture, you know what’s gone on in the claim. This is then fed through AI, it gives a much better predictive capacity, it goes up by about 30%.

”Early in the journey, you can say immediately what should happen with the claim. It helps on the fraud referral and it helps being more effective in triaging,” Thoné added.

With regards to the pandemic, Thoné said this service is needed now more than ever.

Read more…Could Google’s chatbot overhaul the insurance market?

Not subscribed? Become a subscriber and access our premium content

Female chatbot