AWS for Industries

Zero Touch Claims – How P&C insurers can optimize claims processing using AWS AI/ML services

In today’s world, consumers have more power than ever before, and even more choices and experiences. However, this poses a unique challenge for insurers. Typically, insurance companies have limited touchpoints with the customer, given the insurance value chain of quote, buy, and claim. This leaves the provided customer experience as the key competitive differentiator. Simplifying the customer experience is essential for driving higher engagement and meeting the consumer expectations. Furthermore, this becomes even more critical during the claims process. A customer’s experience during the claims process will most likely influence repeat business and determine their long-term retention with the insurer.

It’s no longer enough for insurers to simply optimize manual claims workflows. They must also tackle ever-increasing customer expectations, diverse preferences, and changing demographics to make sure that they’re providing the right claims experience to the right customer. Simultaneously, there’s a competing need to make sure that the claims costs are efficiently managed. Total losses combined with claims processing expenses account for up to 70% of the premium collected. As a result, there’s always pressure on insurance companies to look at newer and more innovative ways of handling and managing claims.

The need to provide both differentiating and personalized claims journeys can be actualized with a combination of Data, Artificial Intelligence, and Machine Learning (AI/ML) technologies. For insurers looking to deliver innovative customer experiences, AI/ML applications can transform the overall claims journey. This allows for both a greater breadth of digital service channels to meet varying customer preferences, as well as the expediting and automation of the entire claims’ workflow – from the first notice of loss all the way through to settlement.

Current approach

Traditionally, claims processing has been a human-intensive process. Insurers have looked to optimize their claims workflows by leveraging robotic process automation (RPA) tools to increase automation levels and reduce manual activities. Although this has delivered some incremental benefits, fundamentally the process remains predominantly manual. In turn, this continues to negatively impact the claim settlement timelines, as well as the insurers’ financial positions, as they must provision for a larger proportion of claims under processing.

Insurance companies have been focused on trying to reduce claims processing costs through automation/RPA and back-office efficiencies. However, when looking to the future, a more wholistic step change is coming, and it will be driven by customer demand. Demographics and customer preferences are shifting with younger generations (X, Y, and Millennials), which constitutes a larger portion of the addressable market. Younger generations show preferences for low-touch self service via digital channels, and they have higher expectations for straight-through processing and overall speed of service.

As insurers look to broaden their target demographics and increase customer retention, there’s a need to differentiate themselves in the claims customer experience area. This will require an analysis of the entire omnichannel customer claims journey, as well as how innovative technologies can be deployed at every phase to provide a seamless and integrated low touch claims experience that aligns with customer preferences.

The opportunity – Zero Touch Claims

‘Zero touch claims’ (ZTC) can be defined as the ability to process the entire claim with no (or minimal) human intervention. Although ZTC isn’t a new concept, it has always been viewed as aspirational by insurers. What has moved ZTC from vision to reality is the ability of AI/ML to replicate complex human-like decision making on various unstructured data, including audio, images, and videos. When deployed in combination, capabilities such as computer vision (CV), predictive analytics, Internet of Things (IoT), intelligent document processing, and natural language processing (NLP) can achieve impressive straight-through processing outcomes.

Achieving a ZTC experience is extremely desirable given the benefits that can be realized for both the insurer and the end customer:

  • Reduced claims operational costs – Staff can focus on higher value customer interactions.
  • Greater consistency and accuracy in claims decisioning through a governed process – Better able to predict claims outcomes early, as well as detect outliers and fraud.
  • Reduced time to settlement – Higher customer satisfaction and retention. Reduced cash reserves for insurers.
  • Innovative and differentiated customer experience – Competitive differentiation

There remains a significant opportunity for insurers to implement AI/ML technologies that drive them toward low and zero touch capabilities. Currently, it’s estimated that approximately 80% of claims go through traditional workflows with human adjudication, with the remaining 20% being auto-adjudicated using some form of automation and business rules. Although there will always be certain use cases where a human interaction is desirable, it’s expected by the year 2030 that the ratios will flip, with 70% of claims falling into the low and zero touch category, and the remaining 30% needing human intervention [1].

Many AWS customers in the insurance industry are already realizing the significant benefits of the application of AWS AI/ML services across the claims lifecycle. FWD is a composite insurer that can straight-through process claims by leveraging chatbots and CV to process images and videos. NIB is an Australian health insurance provider that uses ML-based CV to automate the document processing of invoices and receipts submitted by customers at claim time. Furthermore, Farmers Insurance elevates the customer claims experience with chatbots and conversational AI.

Application of AI/ML in claims processing

For insurers, the claims journey is an area with perhaps the most opportunity for transformation through AI/ML – customer interactions at first notice of loss through to adjudication and final settlement. AI/ML technologies can deliver an improved customer experience through automation and straight-through processing, all while reducing operational costs.

AWS services such as Amazon Lex, Amazon Polly, Amazon Translate, and Amazon Connect help streamline direct customer interactions. Meanwhile, services such as Amazon Comprehend, Amazon Textract, Amazon Rekognition, Amazon SageMaker, and AWS Step Functions help to improve back-office automation and efficiencies. The following figure depicts a reference architecture that details the application of AWS capabilities across each of the core phases of the claims journey.

ZeroTouchClaims diagram

Figure 1: ZTC – Solution Abstract (Click to expand)

First notice of loss

Customers now expect a choice of multiple channels through which they can interact with insurers. For insurers who traditionally have limited customer touch points, the experience provided at lodgement is critical to both attracting and retaining customers. Furthermore, AWS services can be deployed and combined to provide a greater breadth of channels and deliver innovative customer experiences.

first notice of loss

Figure 2: First Notice of Loss – Opportunity Summary

Use cases:

  1. Digital Assistants/Chatbots: Intelligent digital assistants and chatbots have become ubiquitous across industries as an effective and efficient channel through which to service customers. Claims notification/first notice of loss can often be a complex and stressful process for end customers. A guided digital assistant claim experience can improve the accuracy and completeness of the details provided, all while reducing the total time spent by the customer submitting the claim. Moreover, digital assistants eliminate the need for customers to interact with staff via contact centers reducing the overall cost to serve.

Insurers can use the following services:

    1. Amazon Lex – A fully managed AI service with advanced natural language models for building conversational interfaces. By leveraging the same technology as Amazon Alexa, you can build sophisticated voice and text-based interactions. You can also combine it with AWS Lambda[JS1]  for integration with your existing claims, policy, and billing applications.
    2. Amazon Kendra – This intelligent search service is powered by ML. Index your knowledge base, and allow customers to ask natural language questions about claims benefit limits, claim status, and more. This is a great opportunity to ingest product disclosure statements and allow customers to perform Q&A.
    3. Amazon Sumerian – This lets you create and run 3D, augmented reality (AR), and virtual reality (VR) applications. Combine it with Amazon Lex to create a truly immersive and life-like digital assistant claims experience.


  1. Image and Video Capture: The claims process can be greatly accelerated if customers can provide supporting evidence in real-time at the time of notification. The ubiquity of smart phones, drones, and aerial imagery allows insurers to collect high-quality images and video which are foundational in moving toward straight-through claims processing. Image and video collection are applicable broadly across insurance products including Motor, Home/Contents, and Health (receipts/invoices). In addition, cloud storage capabilities are well placed to support the data growth expected in this area.

Insurers can use:

    1. Amazon Simple Storage Service (Amazon S3) – This provides cost effective cloud object storage with exceptional security, scalability, availability, and durability. It’s a great fit for storing high volume image, video, and other documents commonly provided as supporting evidence during a claim.


  1. Embedded devices and IoT: IoT, embedded devices, and smart phones provide a unique opportunity to completely automate the initial first notice of loss. With the increasing number of sensors and connectivity in both cars and consumer electronics, it’s possible to trigger claims workflows with no human input. Devices can stream telematic information in real-time to insurers, which can then be used for claims notification and the optimization of claims processing. In particular, the motor insurance space is the most mature, with many insurers already offering usage-based pricing and discounts to customers who share telematic driving data. The same approach can also be applied to other lines of business.

Insurers can use:

    1. AWS IoT – AWS offers a suite of managed IoT services that can be used to deploy an end-to-end solution that enables devices to stream data for real-time claims, decision making, and analytics.
    2. Amazon S3 – Cloud storage is where customers are building their next generation data lakes. Leverage Amazon S3 to persist all telematic and device data in one location. Native integration with broader AWS data and analytics service offerings can further help in the automation of claims processing.


  1. Contact Center Automation: The advancement of AI-based natural language processing capabilities has enabled the automation of contact centers. Using AI services, insurers can quickly implement smart IVR (interactive voice response), automatically triage incoming calls, understand the customer preferences of self-service versus assisted, understand customer intent, as well as prioritize particular claims related calls based on key words. Customers can start claims-based processes and even ask for claims status updates using voice. The AI-powered automation can scale seamlessly during large-scale events and natural disasters, such as hail storms, earthquakes, and bushfires, where contact center call volumes can’t be handled by human agents.

Insurers can use:

    1. Amazon Connect – An omnichannel ML-powered cloud contact center. Automate claims-based interactions through native integration with Amazon Lex and Lambda. This includes real-time sentiment analysis, transcription, key word alerts, and sensitive data redaction.
    2. Amazon Lex – A fully managed AI service with advanced natural language models for building conversational interfaces. By leveraging the same technology as Amazon Alexa, you can build sophisticated voice and text-based interactions. Combine this with Lambda for integration with your existing claims, policy, and billing applications.
    3. Amazon Polly – Lifelike text to speech capabilities, multi-language, and regional accents.
    4. Amazon Transcribe – A fully managed deep-learning service which uses a process called automatic speech recognition (ASR) to convert speech to text quickly and accurately.


  1. Predictive Analytics: ML models can be trained and deployed to aid in predicting the likely outcome of a submitted claim. This can help across multiple dimensions, including: providing guidance to end customers on expected timeframes until settlement, collecting additional supporting information in real-time to aid adjudication, and prioritizing those claims that require additional attention.

Insurers can use:

    1. Amazon Sagemaker – A fully managed ML service that can be used to build custom ML models to predict claims outcomes and processing time at claim submission. Provides every developer and data scientist with the ability to build, train, and deploy ML models quickly.


Adjudication/loss assessment

For insurers, the loss assessment process is human intensive. Staff conduct manual reviews of claims and supporting documentation, interact with customers (voice + email), determine the extent of loss, and perform data entry. This results in long claims turn-around times, and has a big impact on operational costs. AI/ML technologies can be deployed for both long-tail and short-tail claims to eliminate manual and repetitive tasks and augment claims decision making. In particular, CV (a field of AI) can be deployed to powerful effect in the areas of intelligent document processing, damage detection, and determination.

loss assessment - opportunities

Figure 3. Loss Assessment – Opportunity Summary

Use cases:

  1. Intelligent Document Processing: Claim workflows are document intensive and typically involve multiple exchanges of forms, invoices, receipts, reports, proof of income, and more. This handling and processing of documents in the back office is low value and labor intensive. CV-powered AI services can be leveraged to automate most document processing tasks, including document classification, sentiment analysis, key entity and attribute identification and extraction, form table and handwriting understanding, as well as sensitive information redaction. AI-powered document processing services represent a step change improvement over traditional OCR (Optical Character Recognition) technologies, as they can handle extremely varied input documents (both structure and quality) without needing to develop and annotate templates for every document type.

Insurers can use:

    1. Amazon Textract – An ML-powered document analysis service that detects and extracts printed text, handwriting, structured data such as fields of interest and their values, and tables and forms from images and document scans. It can be used to automatically extract key information from documents and supporting evidence associated with claims
    2. Amazon Comprehend – A natural language processing (NLP) service that uses ML to find meaning and insights in text. It can identify the language of the text, extract key phrases, places, people, brands, or events, understand sentiment about products or services, and identify the main topics from a library of documents. This, in combination with Amazon Textract, can help identify, classify, and prioritize the relevant information in various claims related documents and evidence.


  1. Loss Determination: Images and video have long been used by insurers as primary evidence during loss assessment. The ubiquity and improved image quality of smart phones, drones, and aerial imagery means that CV services can be deployed to both augment and automate loss determination. ML models can be trained to identify objects and determine the degree of damage present across varying property types, including motor vehicles, consumer goods, homes, and more. Example applications include: Identify damaged property in natural disasters using areal imagery, identify motor vehicle make and model, identify motor vehicle damaged panels and the extent of damage, as well as determine if a motor vehicle is a total loss.For low complexity and high confidence use cases, the predictions made by CV capabilities can be used to deliver straight-through processing. For more complex claims use cases, loss predictions can be delivered to augment and speed up the work conducted by human assessors via sophisticated user interfaces and overlays.

Insurers can use:

    1. Amazon Rekognition – This offers pre-trained models and customizable CV capabilities to extract information and insights from your images and videos. Amazon Rekognition Custom Labels lets you identify and label the important objects and scenes in images and videos that are specific to your claims workflows, such as vehicles and parts.
    2. Amazon Sagemaker – A fully-managed service that provides the ability to build, train, and deploy ML models quickly. Leverage SageMaker to automatically label your existing data and build custom sophisticated deep learning models to automate your unique loss determination use cases.


  1. Fraud Detection: Insurers suffer substantial losses (up to 10% of premium) every year due to fraudulent claims. Existing fraud detection solutions tend to rely on rule-based heuristic scoring engines or outsourcing to third parties for manual investigation. As a result, these solutions tend to have a lower successful detection rate and don’t adapt well to changing fraud patterns. ML is well-suited to the fraud detection problem space. ML models can be trained to detect fraud using a combination of large public and private data sets, uncovering contextual relationships and patterns between data that would otherwise be imperceptible. Furthermore, ML models can be easily retrained on new and emerging data sets to continuously optimize detection rates and reduce false positives.

Insurers can use:

    1. Amazon Sagemaker – SageMaker removes the heavy lifting from each step of the ML process to make it easier to develop high quality models. Leverage SageMaker to quickly and easily train and deploy custom trained ML models with your own specific fraud data sets. Proactively detect and alert on suspicious and fraudulent policies and claims.


Settlement and fulfilment

The final phase of the claims journey is settlement. This typically results in either a cash payment, delivery of replacement goods, or a repair. To achieve an end-to-end “zero touch” or straight-through claims experience for a customer in this phase, the focus is on supplementing AI/ML technologies with supply chain optimization, automation, and third-party systems integration.

settlement and fullfiment-opportunities

Figure 4. Settlement and Fulfillment – Opportunity Summary

Use cases:

  1. Orchestration/Process Automation: At settlement, there is a need to execute a number of tasks against disparate systems. Claims systems of record must be updated, payments must be sent to customers and repairers, and supply chain processes must be triggered for replacement parts and goods. Each of these steps can be automated through orchestration services and process automation. If the downstream systems are API-ready, then state machines and orchestration tools are a good choice. For integration with more legacy platforms, including midrange and mainframe, process automation tools should be investigated. The objective here is to completely automate the settlement workflow with human intervention only for exception management.

Insurers can use:

    1. AWS Step Functions – A low-code visual workflow service used to orchestrate AWS services, automate business processes, and build serverless applications. Workflows manage failures, retries, parallelization, service integrations. Utilize Step Functions to orchestrate the overall claims processing flow and integration with downstream systems and third parties.


  1. Multichannel customer engagement: Although called out here in the settlement and fulfillment phase, it’s important to keep customers up-to-date on the status of their claim as it progresses through all stages. Regular status updates reduce the follow up load on contact centers and improves the overall customer experience. Work with customer preferences and deliver personalized updates via the channels of their choosing.

Insurers can use:

    1. Amazon Pinpoint – A flexible and scalable multichannel communication service. Keep customers up-to-date on their claim status with personalized messages delivered via email, SMS, push, or voice. Analytics capabilities let you monitor customer interactions and the effectiveness of different channels.


Implementation approach

Insurers aiming to develop a ZTC experience shouldn’t expect overnight success through large-scale big bang projects. As we have seen, each phase of the claims journey must be examined with the right application of technologies and services to solve each unique use case. Like a jigsaw puzzle, the end goal must be clear and constant, but each step of the claims journey must be tackled independently with it all ultimately combined to deliver a zero-touch outcome.

We recommend starting small by identifying the low hanging fruit. In other words, find the areas of the existing claims journey that have the highest friction with customers, the highest cost, and the biggest impact to claims processing time. In addition, consider starting with low complexity insurance products, and then experiment, test, learn, and iterate before tackling the edge cases and exceptions. Combine this with the tried and tested method of beginning first with the augmentation of human processes. Gather feedback with humans in the loop including end customers, optimize, and then move to fully-automated straight-through processing.

When adopting AI/ML technology, insurers can start with fully-managed, pre-trained services where possible. This will substantially reduce the implementation effort and time to value, all while reducing the in-house AI/ML expertise required to get up and running. For initial forays into AI/ML, it can also be beneficial to leverage the services and experience of specialist partners. AWS has numerous ML competency partners across many geographies who can help you get started.

Although managed software services and third-party specialists can reduce the internal AI/ML expertise required in an organization, it’s still important that broader foundational training and enablement is given proper consideration. The key to realizing the full potential of AI/ML is having internal staff who not only know your business and its challenges intimately, but also understand where and how AI/ML can be deployed to the best effect. Fundamental AI/ML knowledge will become increasingly important at all levels of an organization and will be critical to aligning technology and projects with business outcomes. Explore the extensive free and paid AI/ML training options that AWS has available.


In this post we covered how P&C insurers can leverage AWS AI/ML services to not only automate the claims process, but also achieve the goal of zero-touch claims adjudication. The maturity and accessibility of AI/ML capabilities means that zero-touch claims are now a real possibility for insurers. This promises to deliver a compelling two-fold advantage: innovative customer experiences, and lower business costs.

AWS has the services and mechanisms not only to help build and implement new capabilities quickly, but also to help carriers develop an overall AI/ML strategy. In addition, AWS has the right partner network to support the journey. For further details, contact your Account Manager or AWS Sales Support.

Nick Cuneo

Nick Cuneo

Nick is an Enterprise Solutions Architect at AWS who works closely with Australia’s largest financial services organizations. His previous roles span software engineering, design, and operations. Nick is passionate about security, machine learning, and event driven architectures. Outside of work, he enjoys motorsport and is found most weekends in his garage wrenching on cars.

Gurindar Singh

Gurindar Singh

Gurindar Singh leads Insurance Business Development at Amazon Web Services (AWS) for the Asia Pacific region. He is responsible for leading the development and execution of AWS’s Go-To-Market initiatives across all facets of Insurance (life and non-life insurance, reinsurance and brokers) across the Asia Pacific region. He has close to 22 years of industry experience across insurance carriers and software solution providers and is a regular speaker at FSI events.