Artificial Intelligence
Dec 07, 2018

Applying AI to the insurance claims value chain

The customer journey, data, and claims talent are key to successful projects

Artificial Intelligence (AI) is evolving rapidly across the insurance industry. As the concept and its applications mature, it's already advancing and disrupting how insurers deliver their claims services. The whole claims ecosystem – carriers, third-party administrators, and insurtechs – is beginning to understand the importance of AI in reimagining the customer experience.

Insurers' AI investments

There are many different types of AI - computational linguistics, natural language understanding (NLU), conversational AI, and neural networks – all at different levels of hype and maturity. But AI is already hard at work across the industry:

  • Claims fraud investigations are using AI technologies to trigger fraud alerts, harnessing data from sources such as customer information, external industry data, social media activity, and credit bureau records
  • Claim set-up assistants combine cognitive and NLU capabilities to process unstructured content from emails and attachments at first notice of loss (FNOL) intake 
  • Chatbots using conversational AI in customer service
  • AI models estimating the probability and severity of loss and estimate the cost of auto repairs

These are all improving loss accuracy, shortening cycle time, and freeing up loss-adjusting resources to focus on complex, high-severity losses.

When considering integrating AI into part of the claims process, it's important to assess the business value of each use scenario and not get caught up in the hype of AI. This value can be both financial and non-financial measures, and can provide a roadmap for short, medium, and long-term projects.

As more insurers start their AI journeys, I think there are three key issues to plan for to ensure a successful pilot.

Map the customer journey

Charting the end-to-end customer journey is critical, from both an insured and carrier perspective. Using AI for claims set-up or loss inspection needs to be done in the context of the claims journey to make sure it's joined up with other processes and offers the customer a seamless experience.

For example, a structured triage mechanism built into the claims management platform at FNOL can recommend the best method of inspecting auto damage. This could be routing it to a field inspector or a direct network repair shop, or computer vision-based estimation. Whichever route is chosen, the processes must be in place for a quality review of the estimate once it's submitted, as well as closure of the claim if appropriate.

As the underlying machine-learning engine handles more live claims, its abilities will mature and be able to handle different types and complexities of claims. The underlying claims-operating model and processes must adapt to support this evolution to truly reap AI's benefits.

Humans are the key to AI success

Operational alignment is vital throughout the claims value chain. As AI progresses, claims professionals with reasoning skills and deep domain expertise will play a critical role in making it a success. Immature AI engines will rely on these staff to handle failures, exception paths, and nurture the system to meet its full potential, all within the context of the specific business problem it's trying to solve.

Insurers will need to balance retraining existing claims staff and bringing in the right digital skills to make the project a success. Some insureds will always want an element of personal interaction when making a claim. AI isn't going to replace that interaction; it will only complement it.

Prepare for the data tsunami

AI relies on a solid foundation that needs to be built, kept accurate, and continually developed over a period of time. This includes a multitude of data types, ranging from text (key value data pairs that a linguistics engine relies on) to media (libraries of annotated images, estimate data, and loss reports that are essential for computer vision). My advice here is to start small, in a controlled environment, and focus on achieving positive results and mitigating risk.

But as the pool of data grows, so does the pressure to harness its insights. Blending the AI outputs with other internal and external data sets, like telematics devices, earns richer data and even better insights for action. This means the systems that house the data, as well as the enabling infrastructure need to be customer-centric and scalable to handle the growing myriad of interactions across the life cycle.

About the author

Girish Malik

Girish Malik

Vice president, Insurance practice

Follow Girish Malik on LinkedIn