AI in claims

The implications and opportunities for the London market, personal and commercial lines

In preparation for London Market Claims, we interviewed Paul Willoughby & James Russell (Folio group’s claims subject matter experts) on their perceptions of AI and it’s uses in the claims arena. Paul answered from a London market perspective and James gave the Personal/Commercial lines perspective.

Can you give some tangible reasons why companies should be looking to experiment with AI, specifically in the claims arena?

Paul: I think that there is a real lack of investment in claim technology (especially in the Specialist space), businesses tend to only upgrade their claims systems every 10-15 years, when they do they get the benefit that these technologies have the capability to automate tasks and improve important things like triage etc, but those with legacy platforms can’t benefit from these advances in technology. I think there is a huge opportunity to experiment on things like data extraction, claims data ingestion and claims triage outside of the legacy platforms and then feed the data into the strategic platforms. These experiments and the resulting benefits will feed the return on investment model to justify a larger change programme, e.g. if we have saved X in Y months with this small investment, let’s look at the larger opportunity.

James:  The immediate reason is because AI can do what humans don't have the time or capacity to do.  Sure, people could look at each claim in detail for all the background on parties involved, circumstances etc.  People could work out optimal settlement strategies - but the reality is that claims teams don't have the resources to do this cost-effectively on every claim.  So AI can help automate and augment the work of the claims handler.  Machine learning models can help us understand the parties involved, what their needs and preferences are, identify potentially vulnerable characteristics, and suggest optimal settlement strategies.  Claims can then be automatically triaged and routed:  Simple claims can be straight-through-processed with no human intervention needed.  Human claims specialists are then freed up to focus on more complex claims where the customer needs more support, or the settlement strategy isn't so clear to the AI. 

 

In what ways could AI-driven claims automation shift the balance between efficiency and empathy, is this a good thing or bad thing?

Paul: I’ve worked with some of the best specialist claims experts in the world, when they get a chance to react to a situation they are absolutely brilliant. The feedback from Risk Managers and end clients always reflects their expertise and passion, and the fact that they went over and above to manage the situation.  Now obviously there are not enough experts to sit across every claim or incident, so you have to be very picky about when to deploy an expert and when you use an automated solution.  The variables change depending on the segment of the market but in essence we kind of know when a human needs our attention after an incident and when they want to be left alone and to just receive a cheque.

I’ve had this concept of an empathy bot for a while, i.e. an AI that can look at the initial incident and detect what has happened based upon past examples.  A great example may be a commercial fire loss where the client has purchased business interruption insurance, that’s going to be a very different conversation compared to someone that doesn’t have that cover.  There are going to be different stress levels and different levels of managing expectations required. What we really need to be investing in is automation tools that can detect these examples quickly and push the red button on claims where a human touch is needed (or expected).

James: Efficiency and empathy can't be mutually exclusive.  We have to do both as well as we possibly can.  I agree claims has focussed too much on efficiency in the past and if we are to meet customer expectations and reverse the trend in consumer satisfaction, we need to be empathetic as well.  In a previous company we once used the maxim "Get me, get the plot, get it done".  AI, supporting the human can definitely help us handle claims efficiently and empathetically.  So I'm going to use this maxim to make my point for a damaged private car scenario:

• 'Get me' - From multiple data sources I can see that this customer has had this car for a while, it has a cherished plate, it may have sentimental value
• 'Get the plot' - The reality is new parts are hard to get for this car and repair estimates based on image damage assessments indicates it's beyond economic repair for us as an Insurer
• 'Get it done' - Best course of action is to check how the customer would feel about a more cost effective repair using recycled parts or a cash settlement for them to retain the vehicle and arrange repair themselves 

Data and AI models can do a lot of the leg-work for a claims specialist.  If empathy is about acknowledging someone's circumstances and reacting appropriately, then AI can definitely help us be efficient in 'get me' and 'get the plot' stages.  If this can then stop a vehicle being forced down a repair or total loss route which is likely to lead to friction and customer dissatisfaction later on because a cherished plate has been disposed of with the vehicle, then we are both being efficient and empathetic to the customer.  That has to be a good thing.

 

How do you recommend companies get started?

James: Simplify.  Get back to basics.  'Systems Thinking' is a really powerful place to start.  Go back to what really matters to a customer and what we need to achieve when someone suffers a loss and makes a claim.  Focus on activities that add value to settling the claim and eliminate effort that gets in the way or doesn't help.  Start with simple data visualisation and propensity models that can surface insight that humans don't have time to find or work out.  Simple dashboards that sit alongside established claims admin' systems can be inexpensive, provide early value and enable progression from basic data insights to AI powered 'next best action' recommendations.  This learning can then feed automated routing of a claim and automated settlement.  Time is saved, accuracy can be increased and value generated to justify further investment.  Above all - BRING YOUR PEOPLE WITH YOU!  Technology and AI is only half the solution.  You won't maximise efficiency or empathy without people working alongside the AI.

Paul: I think insurers are businesses built on operational processes. We layer on more and more processes as and when the regulatory model changes. Lots of people are looking at AI as a huge opportunity to transform what they do, but what they are forgetting is the marginal gains i.e. making small, incremental improvements in various aspects of a process or system, which cumulatively lead to significant overall enhancement in performance or outcomes. If everybody within an organisation decided on one process or task that took more than 10 minutes but the human element added little or no value then cumulatively you would be saving days and weeks, and this doesn’t even take into account the impact it may have on staff morale, culture or client satisfaction.

I’m working with clients at the moment and we’re prioritising lots of small experiments that when combined will have a huge impact on the operating efficiency of the organisation, but most importantly it’s going to free up the practitioners to work more closely with clients and do what they were brought in to do.