Man Vs. Machine
By: Anthony Harrington
There are two diametrically opposed views of where claims automation in the insurance industry, generally, is going. On the one hand, there are already press headlines about artificial intelligence (AI) and software robots taking over claims processing, leading to mass redundancies among insurers’ claims staff. The alternative view is that claims automation, even ‘smart’ AI-driven claims automation, will simply digitise the boring, process driven, ‘if-that-then-this’ elements, leaving experienced claims staff with far more time to either engage directly with customers, or put more time into examining potentially dodgy claims.
Of course, in many ways, this is simply the insurance industry’s version of the long running debate about whether technology is enhancing our lives or whether the paranoia about robots taking over the world has something going for it after all. In a very real sense, however, the debate is meaningless, for in the view of many experts, the insurance industry has no option but to embrace digitisation and automation just as fast as it can.
The crunch point is that customers want to be able to deal with the industry via their smartphones or the device of their choosing and insurers without an effective, efficient digital channel will be at a clear disadvantage vis-à-vis the competition. Moreover, because a fairly high degree of automation is part of what makes a digital channel fast, efficient and effective, automation is an unavoidable, inevitable part of this process – though each insurer will have to decide for itself where the hand over between a robotic system and human experts should happen (or even whether it wants to keep humans in the loop at all!).
Jamie Hersant, head of lifestyle claims at AXA, makes the point succinctly: “As insurers, we have to evolve solutions that deliver what the customer wants. So there is no option here for any insurer but to move forward and embrace both digitisation and claims automation,” he says.
Customers want to do things in their own way, and in their own time. So, in terms of travel insurance, this means automating virtually all straightforward claims, such as claims for lost baggage, lost mobile phones and protracted flight delays or cancellations. Judging whether a large medical emergency claim is excessive or fraudulent is another matter entirely and there is still considerable doubt over just how far automation can go in accurately deciding complex ‘big ticket’ claims.
Hersant points out that, as one of the largest insurers, AXA was one of the first to invest significantly in online claims processing. The next step, which the company is close to announcing, he says, will allow customers to track the progress of a claim online, using whatever device they choose. “Our systems are ready and able to deal with this challenge,” he says.
However, full automation, which would entail AI ‘robots’ making complex decisions about whether a particular claim is fraudulent or overstated, still seems some way off. Most of the industry (though not all, as we will see) still believes that expert assessors are much better at this than any AI driven process. Hersant explains that AXA has chosen to automate the process of submitting the claim, while keeping the expert assessor very much in the loop as far as making decisions about the validity of larger value claims is concerned. At the same time, AXA has built software solutions as part of the claims validation process that zero in on common risk factors in higher value claims and can flag up a particular claim for further examination by an expert assessor.
“When a customer builds their claim online, our system works out a risk profile for that particular claim, and if it hits certain key ‘triggers’ then the claim is routed, via our automation and workflow software, to an expert for further examination. In this way, the technology supports our claims handlers in a way that we weren’t able to do prior to automation,” he comments.
Of course, it is always possible for someone to attempt to systematise the kinds of analysis and reasoning that a claims expert goes through when deciding whether and to what degree a particular claim needs to be investigated. This is exactly what Fukoku Mutual Life Insurance is doing. According to press reports, the insurer expects to lay off some 30 staff and save around £1 million a year once its £1.4-million AI claims automation system is installed. Based on IBM’s Watson Explorer AI, which IBM claims can analyse unstructured text, images, audio and video, the system is going to be used to calculate the 132,000 payouts to policyholders the insurer makes in a year. However, Fukoku Mutual will still have a human expert in the loop, since the final decision on any payout will still be made by a member of staff. And, of course, the insurer will continue to require expert claims investigators to look into any dubious claims.
Streamlining global payments
So far, we have been talking as if automating anything short of an investigation into an obviously dodgy claim is a straightforward process. In reality, solving the various challenges involved in ‘straight through processing’, which would see a small claim, say, processed and the money automatically deposited into the claimant’s account, can be non-trivial. Anil Sawrup, chief commercial officer at Cambridge Global Payments, points out that the routing instructions to get payments into client bank accounts can be very complex, particularly where the claimant is not resident in the same country as the insurer. So the insurer might have a perfectly adequate claims automation process, but they will still need a payments capability to close the loop on the claim by ensuring that the right payment gets through safely to the right bank account.
“Take Brazil, for example. There is a lot of complexity involved in making sure that you have all the necessary banking information and all the ancillary information that is needed to get payments progressed through to claimants’ accounts,” Sawrup notes. Any country that is engaged in tight management of its currency, perhaps to control marked imbalances in the position of its currency versus foreign currencies like the dollar, euro or sterling, is going to have complex rules relating to currency payments into local bank accounts. “What we do is all the upfront screening and validation, particularly relating to the complexities introduced by various national anti-money laundering provisions. We also screen all the details relating to the beneficiary, making sure that all the data is clean and in order before we release the funds for payment,” he says.
A key part of closing the payments loop is to have a very secure in-country messaging capability, and the ability to customise the payments technology to interface with the insurer’s systems. “Our approach is very much a cloud-based system with no software installation required. It is all web-based, so the company that we are selling our services to does not need to make any upfront capital expenditure. We do all the data mapping to get the information from their system into the payments system free of charge, including cleaning and validating beneficiary information,” he comments.
“It is important to stress that we do not validate claims. That is down to the insurance company. What we do is to enable them to pay out claims in a secure way, without them having to solve all the messaging, routing and regulatory requirements involved in making both local and cross-border payments,” he says.
This becomes particularly important when an insurer is trying to open up a new overseas market for one or more of their insurance products. Claims for payment are a natural corollary of any insurance product, and handling these payments in foreign parts requires specialist payments technology to complete any automation process.
Making data work
Another piece of the technology puzzle that insurers have to solve is how to get paper- based correspondence into digital form so it can be handled by their claims automation system. Lexmark’s Jeff Hiegert points out that document scanning, document management and workflow are all things that companies such as his can bring to the table to complement and complete claims automation. Hiegert has spent the past five years working for Lexmark and the past 20 working in various capacities in the insurance sector, as a property casualty underwriter and in various insurance vertical roles. “Over this entire period, I have seen insurers working towards digital solutions, trying to get as digital as they can to bring benefits to customers and efficiency gains to themselves,” he comments.
One of the biggest parts of this push to digitalisation has been the effort to translate stagnant paper-based data into digital form so that it can be mined for ‘big data analysis’, showing trends in claims, fraud and so on. “Data is ‘the currency’ of insurance and the challenge has always been to make this data work. Insurers have vast amounts of data that could help them but until recently they have just had no way to sift it. Time and again, one saw insurance companies go into projects designed to improve their operations and these would fail because they were working off false assumptions. The challenge has been how to get that data off the paper and make sense of it all so as to make more informed decisions,” Hiegert says.
What Lexmark does is to create what it calls an ‘omni-channel’ experience that can collect and direct communications and information from all sources, email, online and paper, into the claims automation system. “We try to break any business process down into five phases: initiation, collection, evaluation (or decision making), interaction and then closure. We can then look at the various ways an insurer engages with the customer or the internal stakeholders of the process at each of these phases. That enables us to find opportunities for digitisation and automation at a fine level of detail,” he explains.
One of the smart things that companies such as this brings is the ability to recognise the type and purpose of a paper form. It can tell if it is digitising a claims form, versus an application for travel insurance, for example. “We can identify and classify the type of document, extract the relevant data off the paper and feed it into the claims process. Business rules can then be applied to automate routing, trigger certain investigation tasks or even submit the document for straight-through processing,” he says.
The role of more sophisticated AI in claims automation is still in the experimental stage, he believes. Hiegert recently gave a presentation on cognitive computing at a conference on the Internet of Things and argues that there are two fundamentally different types of AI being talked about, namely ‘doing’ and ‘learning’. The roboticisation (or doing) of simpler processes and tasks is going ahead full speed. The learning part is really about data analytics, or pattern recognition operating on historic data in big data warehouses, which is useful mainly for trend analysis on claims. While insurers seem willing to let technology handle some of the ‘doing’, they still want human experts to play a significant role in the ‘learning’.
“What we see is insurance companies using AI to mine data and then using human logic to check the validity and usefulness of what is coming out of the patterns. For decisions about complex higher value claims, it is still very much a question of just routing that claim to the relevant expert,” he concludes.
Read the full article here: https://www.itij.com/feature/man-vs-machine