Harnessing intelligent use of data to revolutionize claims processing

Manual data entry errors, lengthy processing times, and rises in fraudulent claims have long hampered insurers’ efforts to meet customer demands.

However, artificial intelligence (AI) allied to advanced analytics and machine learning are now turbocharging and revolutionising claims processing.

Data Entry

Automated data entry systems reduce the errors that can occur in multiple entries. For example, the automation required to comply with the London market’s Electronic Claim File system cuts down the risk of mistakes slipping into the first notice of loss (FNOL) stage.

These smart systems can manage a client’s inbox and populate all the required fields rather than have a human read the initial claim email to work out what happened and when, then key in the pertinent points.

These systems can read emails and attachments, look at images, and create a file – all at a fraction of the time it would take a human. Only then does an individual look at the file and make corrections where the AI may have misinterpreted data. But even if there is an error rate of 5% within the automated data, it is still quicker to correct those errors than write the whole report from scratch.

While these automated data entry systems bring efficiencies by reducing the burden of administrative tasks, their ability to learn is also a significant advantage. For example, in the United States, dates are entered with the month then the day – the 15^th of January is written 1/15 in the US - but if the system doesn’t recognise the American format, it can be trained to understand and accept it.

Missing Data

These systems can also be trained to spot any gaps in the claims forms and automatically request the missing information from the sender, allowing a much quicker transition to the next stage of the claims process.

Take, for example, a flood claim. Once all the data is in the system, AI tools can be used to interrogate the information and deliver a report about what happened, what the damage was, and the replacement or repair value. It can also check historic losses from an enormous database of claims to check if a similar incident has happened before, and how much was paid out.

The automated data entry systems can revolutionise claims handling at the FNOL stage, significantly reducing data entry time by weeks or even months, and saving around an hour in processing time, which can really add up for insurers which are processing tens of thousands of claims annually. Having these automated systems also means insurance companies can rely on relatively small groups of skilled insurance professionals who can use their experience and market knowledge to carry out value-added oversight and quality control tasks.

Making the Best Use of Automation

Automation is currently most appropriate for use in a straightforward claim and in the early stage of data entry and interrogation rather than in the complete end-to-end process.

Some insurers may want to automate the “safer” parts of the process such as settlement and fraud questions, whilst others may want to automate more of the process.

However, it is not appropriate to automate ‘sticky’ claims that are difficult to resolve and may drag on for a considerable time, which might include legal disputes or protracted negotiations. These complex claims require a high level of expertise and judgement by the likes of senior claims handlers.

While automated data entry systems reduce the amount of administration humans have to do, allowing them to focus on more high-level tasks, there are risks involved if the data is not reliable or there is a limited dataset on prior claims history. For example, if there are only a hundred or so claims of a certain type in a specific year, it may be difficult for AI to detect patterns or trends than if it was interrogating a much larger pool of prior claims.

The quality of the data is key to the entire insurance process, which underscores the need for insurance companies to have robust, reliable and comprehensive databases if they are to realise the full benefits of automation.

Human Intervention

Human intervention is still essential, however. For example, an insurer may have clients moving goods from Spain to France, and a lorry overturns in France. The AI would record this as a French claim, when from a claim ownership point of view the client might want this recorded as a Spanish loss. That is why it is vital to always have human professionals analysing the data.

This type of example also underlines the importance of not using AI to automate the whole process, but rather to focus on the easier, labour-intensive tasks whilst expert humans oversee the process, certainly for the foreseeable future.

Fraud Detection

Today, algorithms also make fraud detection easier. Fraud tends to follow patterns, and automated systems excel at detecting these trends. For example, multiple claims from the same claimant would trigger an alert. Other red flags which a programme could detect could include claims happening close to the policy start date, uncommon types of claims, or instances when claimants refuse the use of an insurer’s repair network.

Risk Management

AI can also be used intelligently to add value to risk management and prevent losses occurring in the first place. For example, a delivery company consistently had fewer vehicle losses in the UK than in France.

The data revealed that the UK vehicles were fitted with 360-degree cameras, whereas in France they were not, prompting UK drivers to take more care as they were being recorded. The camera footage also provided more claims information to determine who was at fault and assist in recoveries. Risk management teams can then present their CFO with the cost of claims compared to the cost of fitting cameras.

Many third-party administrators (TPAs) simply deal with the fallout from a claim, but clients would really benefit if the TPAs also offered value-added risk management advice.

In the case of the French firm, they could advise them that if they decided to fit cameras to their vehicles, it could lead to less accidents, reduced driver injuries (time off work), a reduction in reputational damage, and reduced premium saving them money in the long run. In addition, if the cameras then recorded the causes of accidents, the company could take further steps to prevent these incidents through additional training, improved equipment or delivery routes.

Intelligent claims processes

Intelligent data solutions combined with advanced analytics and machine learning unquestionably reduce the administrative burden, speed up the process and make it more efficient and less error prone.

By properly harnessing data, insurers can supercharge efficiency, cut costs, and enhance the customer experience.

While automation will undoubtedly bring significant benefits to the claims process, this will not lead to the elimination of human beings, who will continue to play a vital oversight role.