AI in the Insurance Value Chain

 

Technology advancements and customer expectations are rapidly changing traditional industry boundaries. Platforms that connect offerings and ecosystems that connect services have emerged across industries.

 

Insurers can embed their products into seamless customer journeys by tapping into an ecosystem. We live in an interconnected world where embracing ecosystems is crucial to reaching customers at the moment of need, whether that’s fostering direct customer relationships or integrating with organizations that own the customer interface.

Expectations of clients are as high as ever, and changing rapidly with younger generations buying insurance for the first time – digital channels are a must. 

 

Insurance products and services require adaptation to meet these needs. Investing in customer-facing, cloud-based digital innovations and building new ecosystems that meet a wide range of these needs is only part of the solution. Traditional back office processes need to be refreshed for efficiency and compliance as well. 

 

The industry is at a pivotal point, and now’s the time to provide meaningful value to customers, employees and stakeholders.

 

Artificial Intelligence and Intelligent Automation, with application of technologies like Machine Learning or Natural Language Processing, give insurers a valuable competitive edge. In this article, we will support this with a few examples of applications of AI in Insurance, along its value chain.

 

 

Insurance companies classically rely on historical loss data in their actuarial models in order to assess and price risks. The challenge is that in emerging markets and emerging risks, that data is often not available.

 

Insurance underwriting has always been about including reliable and available data sources to come up with a risk and pricing for an insurance contract.

 

With the availability of more data sources and the sheer processing power to include all of those in a better model, machine augmented underwriting is the new normal. To keep a competitive edge, it is key for modern underwriting to use AI in insurance trough machine augmented features to keep up or be ahead of the competition.

 

This is why some Reinsurance companies like Munich Re or Swiss Re provide such solutions to direct insurers, in order to enable them with the metrics that matter to them, not only relying on historical data, but also on other available data sources.

 

With more and better data, introduction of descriptive, predictive and prescriptive analytics based on machine learning or other modern techniques, can be of considerable value to insurance carriers.

 

Financial Services is still a very much paper based business. An offer to accept credit conditions, a credit contract, a bank account contract, forms for activating online banking, forms for changing your address, phone number. Bank statements are sent on (guess what) paper.

 

On top of that, emails are exchanged, service phone calls are made, online chat, etc.

 

Most of the data is unstructured and archived in either physical or digital archives as text documents. In insurance, text documents constitute 80 to 90% of the overall data. This data is often not accessible, not even searchable.

 

The opportunity to use Natural Language Processing to unearth this enormous value presents itself. Using techniques from text disambiguation, to automated classification, entity extraction and data linking, enormous value can be extracted from unstructured source, helping optimizing processes, revealing new insights, and even evaluating risk portfolios.

 

 

Due to legacy insurers’ reliance on paper-based forms, there is an opportunity to automate processes based on paper documents.

 

• Optical Character Recognition (OCR) – a tech-enabled method of recognising text in scanned documents. OCR processing of all incoming paper documents is the very first step that needs to be done.
• Classification and Routing – through automated classification of documents (e.g. invoices, claims). Now, an intelligent system can be trained to automatically classifying documents and routing them to the right departments and workflows.
• Extraction of data – next, depending on the type of document different types of information can be extracted (e.g. date of loss, client number) in order to prepare for the next step in the process.

 

The process above can be called Document Capture and is typically integrated with an Enterprise Document Management System (EDMS) and a Records compliant Software.

 

 

For repetitive, standardized, and attention-demanding workflows, intelligent automation may deliver great ROI. Claim management is a great example.

 

Most of the claims management process is paper-based, and it rarely receives end-to-end digitization. As a result, it can consume up to 80% of premium income. The manual nature of claims processing also leads to errors and inefficiencies, which increase insurers’ operating costs. By 2019 larger insurance carriers had not yet adequately addressed the cost of services delivery, as stated by McKinsey at the beginning of 2019.

In 2022 many insurance companies plan to achieve greater operational efficiency with the help of new technologies, including:

 

• AI (machine learning, deep learning)
• RPA (robotic process automation)
• IoT (internet of things).

 

Thanks to the increase in connectivity – telematics and onboard computers in cars, smart home assistants, fitness trackers, and other IoT devices – insurers can now collect more information about their customers than ever before.

 

Then they can use it for underwriting and claims management tasks to make them faster, more agnostic, and less error-prone.

 

Data enables better decision-making and reduces risks. Nevertheless, larger volumes of data require more advanced (and secure) processing methods. Artificial intelligence algorithms come into play here. The algorithms are able to effectively scan all the incoming data, interpret it instead of insurance agents, and provide faster settlements.

 

Training data allows machine learning and deep learning algorithms to improve over time without explicit programming, giving your teams access to even more accurate and complex data.

 

Here are some AI use cases in claims management:

 

• Initial claims routing
• Claims triage
• Fraudulent claims detection
• Claims management audit

 

Fraud is a growing problem, particularly in the fields of motor and P&C, and insurers can gain considerable profits and savings by fighting fraud. The time is now for insurers to act.

 

According to this McKinsey article, in 2015, there was plenty to be done in terms of approaching Insurance Fraud detection in an holistic way. The main issues identified by their benchmark were:

 

• Not in the focus of top management.
• Limited importance of fraud in operational claims processing.
• Insufficient specialization.
• Hardly any modern investigation methods.
• IT systems that are obsolete or have not been maintained.

 

So, Insurance companies need to approach the topic of Fraud Management in a more holistic way and develop frameworks for fraud benchmarking and the definition of a fraud process and funnel.

 

After having introduced a process, the same can be instrumented and digitally mirrored, so that insights and optimizations can be done.

 

The recipe for introducing AI in such a process can look like the following:

 

1. Formalize your fraud process.
2. Instrument it. Introduce measurements and collect data about every step. The data that comes in and out of each stage, what people have done and when (aka collecting annotated data). 
3. Look for potential spots for efficiency improvement
4. Introduce automation using Machine Learning (or not, sometimes there are simpler approaches)
5. Iterate

 

 

The path to modernisation, digitalisation and implementation of AI in insurance companies is not always straightforward, mainly due to the required changes in the industry’s mindset and strong regulation. It is key that insurance companies take the journey with the right technological partners.

 

If you would like to know more, check our media section and watch or listen the podcast episode about Automating Insurance Processes with AI.