Q&A: Shift Technology co-founder and chief scientific officer Eric Sibony on the future of automation and ‘autonomy’ in claims handling

What would you describe as ‘true’ claims automation? And how does it marry up with buzz words that are increasingly bandied about like ‘low / no touch claims’ and ‘straight through processing’?

True claims automation is an opportunity for claim handlers to offer the service and guidance that they don’t usually have enough time for, instead of being hampered by all the tedious administrative work.

From the insurer’s point of view, ‘no touch’ is where the claimant never has any interaction with a claim handler, face-to-face, on the phone, or via email. The claimant might have interactions with others, if they have to take a vehicle for a repair, or have a repair done in their home. ‘Low touch’ signifies a minimal number of interactions with a claim handler. Straight through processing is where the process is adapted and fast-tracked, which in some cases means dropping some of the controls for certain types of claims. This is about process rather than touches, but it is all part of claims automation.

Carrying on the theme of ‘buzzwords’ what is the difference between Robotic Process Automation and Artificial Intelligence and how they can be used in automating claims?

For the sake of an analogy, automated trains have been around for 10 years, while driverless vehicles are a much newer example of automation. We don’t call them automated vehicles; we call them autonomous vehicles, which is a hint about the complex tasks they perform. A train just goes in a straight line and the automation there is just about when to stop and when to open the doors. It’s very basic. For example, it doesn’t have to be able to increase or decrease speed. It’s much more complex to build an autonomous vehicle because it must be able to react to the unexpected. RPA is the equivalent of automated trains; AI expands the possibilities at the level of driverless vehicles. There are some ways to apply AI, where you can get to that kind of ‘autonomy’ I would say, rather than automation.

What are the benefits for customers in insurers automating their claims journeys?

People talk a great deal about the time of claims. But in many claims, at some point, there will have to be a cost estimate and repairs. Automation doesn’t change that. So, you can only gain a few days in those cases, which will last for a few weeks.

I think the main benefit is the simplicity you get with automation. Claimants benefit from real-time resolutions and answers on digital channels whenever they need it. So, they don’t have to call, which can be quite time-consuming and inconvenient. And there’s this new kind of concept, which is ‘first time resolution’ where insurers identify claims that can go straight through and try to solve the claim at time of first notification of loss. All of this is leading towards a much simpler customer experience.

What about customers who want to interact with their insurer/speak to someone; how do you marry that up with claims automation?

Automation is even better for customers who want to speak to someone on the phone, as claim handlers can provide better service to them. Claim handlers can be there to provide guidance in complex cases. And the claimants who are at ease with the technology won’t take up their time. So, they can devote their time to the people who need it. And in any automated process, whatever the application or digital channel, the claimant should always be able to ask to be called. Conversely, when the back office is automated, the claim handlers should be able to intervene at certain points if necessary and call the claimant.

Can all claims be automated, even complex, high value ones?

Things that happen outside of the insurer, like repairs or the involvement of medical providers, is out of the scope of automation that we are talking about. Maybe one day in the future, there will be drones to fix a claimant’s roof. At the moment, this part is not automated as somebody needs to do that.

On the insurance side, claim handlers are only doing a few things. They’re processing the information they get from the claimants and then they have to give an answer. They are also coordinating and information sharing between stakeholders. If you look at it in those terms, there’s no theoretical limit to what can be automated by AI. Because even with complex claims you might have some small parts that can be easily handled. This will happen progressively, starting with simple claims or with the simple parts of complex claims. But in the long run, there is no reason why some claims would necessarily require a human. Although some customers might choose to have a human.

We have talked about customer benefits of claims automation, what about the internal benefits for staff?

The obvious benefit is a reduction in workload and an increase in productivity. Claim handlers may not want automation to do everything, but it helps by doing certain tasks in advance. In many claims for example, for a subrogation or legal or other type of process, a claim handler will need to write something and send it by post. This can be quite tedious because it requires them to synthesise a lot of information. Automation can help with this through an application a bit like Gmail’s predictive sentences, but with a broader scope. This kind of feature makes these tasks faster and easier for claim handlers and helps them avoid mistakes. So, another benefit from the insurer’s point of view is the reduction of errors and leakage.

Continuing this inward focus, what are the common challenges – both in terms of culture and technology/data to overcome in automating claims – and how can they be overcome?

I don’t think there is a major challenge in culture. The claim handlers and claim managers for the many insurers we have worked with are keen on claims automation. And the tendency for customers to call instead of use digital channels should change progressively.

Technology remains the biggest challenge. Because insurance has become more and more sophisticated. The complexity this entails is most easily handled by human beings with life experience. For example, some insurance products are there to cover the fees for repairs based on some specific circumstances in car accidents or with water damage.

Now everybody has been in a car or a house. But an AI is starting from scratch, having never lived in a house and not even knowing what water is. Claim handling rules are quite straightforward for humans to apply. For example, if the water came in through this kind of window, and the window was open for this kind of reason, then it is covered or not covered under the policy. For an AI this is very hard, because the AI doesn’t have the background knowledge. AI has made huge progress, but we’re not there yet. This is the technical challenge we are battling with our solution. To get to a full and exhaustive automation or autonomy, there is still some work and R&D to do in that direction.

The same problem arises in terms of unstructured data, including pictures, documents, and text, including voice translations. There have been some significant advancements in all these areas. But again, an AI lacks the life experience to be able to analyse pictures, categorise documents and understand conversations.

Finally, one criticism that is often mooted when claims automation is mentioned is that it increases an insurer’s vulnerability to fraud by removing the human touch; how would you counter that argument?

Automation [in claims] must come with automated fraud detection. Shift Technology actually started here and then expanded to automated claims. And that is the right way round. You cannot have claims automation without fraud detection. And the fraud detection has to be real-time. And with automation becoming more and more common, it must also have the capacity to learn and adapt and to catch fraudsters even as they evolve and develop new ways to commit fraud.

Sponsored