THE COMPUTER VISION ADVANTAGE FOR INSURANCE CLAIMS E-BOOK

Transcription

1 THE COMPUTER VISION ADVANTAGE FOR INSURANCE CLAIMS E-BOOK

2 Table of contents The multiple challenges facing the insurance industry 3 AI Embedded in Insurance Processes 5 The Rise of Computer Vision 5 Examples of computer vision technologies 6 Augmented Reality s Link to Computer Vision 9 Computer Vision s Potential in P&C Insurance 10 Computer Vision in Customer Service 11 Assisted Service 12 Self-Service 13 The Synergy between Assisted and Self Service 14 Examples of the Synergy 14 Potential Insurance Use Cases 15 The Value Computer Vision Brings to Insurers 17 The Challenges of Implementation 19 How to get started implementing Computer Vision in insurance 20 Summary 22

3 The multiple challenges facing the insurance industry Within the global insurance industry, success is often measured by the efficiency of the claims handling process. Financially, claims represent the largest single cost to insurers with more than $2.29 trillion in claims paid by in 2016 alone, and are considered a critical operational factor for property and casualty (P&C) insurers. In addition, both the efficiency and ease of the claims process have a significant effect on customer satisfaction. Insurance companies are seeking innovative ways to differentiate their offerings from the competition and be more customer-centric in order to gain a competitive advantage. P&C insurers are increasingly automating processes by leveraging technologies such as artificial intelligence (AI) and robotic process automation (RPA). This has given rise to a new era of Insurtech companies, new players using technology to disrupt the insurance industry with innovative business models that focus on a fast and efficient approval process. In addition, for the P&C industry where fraud is an $80B per year problem, implementing 3

4 digital functions based on data and analytics is necessary to both enable an elevated claims experience for customers, as well as to boost loss prevention and overall profitability. Companies that strategize their efforts to improve the insurance claims process can expect higher levels of operational efficiency, as well as a greater financial impact and more satisfied customers. While each claim is different, the standardized elements of how quickly claims are handled can be improved by implementing better insurance processes and utilizing technology more strategically. 4

5 AI Embedded in Insurance Processes Multitudes of claims, customer inquiries and masses of data make the insurance industry an ideal address for artificial intelligence and cognitive technologies. A 2017 study reports that the insurance industry has invested $124 million in AI, compared to an average of $70 million invested by other industries. According to McKinsey s Insurance 2030 Report, with the new wave of deep learning techniques, such as convolutional neural networks, AI can truly mimic the perception, reasoning, learning, and problem solving of the human mind. AI will help the insurance sector shift from its current state of detect and repair to predict and prevent. By embedding AI within claims processes, insurers can easily access and extract the relevant data and reduce processing times. AI can also identify patterns in masses of data and help detect fraudulent claims during the process. Using machine learning capabilities, AI can automatically calculate past damages and predict the costs from historical data, aiding in the determination of premiums. McKinsey predicts that as insurers and their ecosystem become more adept at using AI technologies to enhance decision making and operations, lower costs, and optimize the customer experience, the pace of adoption will further accelerate. While AI has made great strides within the insurance industry, one promising element of a robust AI strategy must be further explored: computer vision technologies. The Rise of Computer Vision Computer vision is the science that deals with enabling computers to see, identify and process images in the same way that human vision can. Via automatic extraction and analysis, computer vision enables the machine to extract meaningful information directly from an image, and then utilize learned algorithms to achieve automatic visual understanding. Computer vision is based strongly on artificial intelligence, as the computer must gain high-level understanding from digital images or videos, and then perform appropriate analysis or act accordingly. Since 2015, computer vision has become more mature, with image recognition achieving significant improvements thanks to deep learning techniques, and in some fields even surpassing humans in the ability to recognize objects. Computer vision technologies can add significant value to the insurance sector by giving eyes to agents, collectors, underwriters, assessors and customers. 5

6 Examples of computer vision technologies Examples of computer vision technologies that bring value to the industry are: OBJECT RECOGNITION Enables finding and recognizing objects within images or videos. Object recognition includes several tasks, such as: classifying objects, localizing the object within the image, distinguishing the object from other objects, and identifying parts within the object. Deep learning-based object recognition offers incredible accuracy that makes object recognition a core technology for many insurance-based applications, including onboarding, underwriting, claims assessment and damage recognition. IMAGE TO TEXT Technology provides textual descriptions of the object or issue it identifies within an image, using deep semantic alignment. A machine develops this capability by recognizing objects and their locations within an image, converting this information into text, and creating a meaningful contextual sentence to describe the image. This technique may be used by insurance companies to help guide customers through the image capturing process, with IoT device onboarding or with billing/contracting issues. FACIAL RECOGNITION Deep learning has made significant improvements to machines facial recognition capabilities, especially in challenging lighting conditions, angles, and backgrounds. With facial recognition, insurance companies can automatically identify and authenticate customers, speeding up the validation process and reducing the risk of fraud. 6

7 IMAGE SIMILARITY Is a technique used by a machine to analyze an image and search for similar images within massive visual data sets of captured cases. TEXTURE RECOGNITION Is the process by which machines understand, model and process texture. Texture analysis may be used by insurance companies to identify material-based objects or to classify damages to carpets or other types of upholstery. SCENE RECONSTRUCTION Refers to the process of capturing the shape and appearance of real objects in 3D. This technique can be applied to reconstruction of an accident scene or measuring damage to a home. MOTION ESTIMATION Uses algorithms to determine motion vectors that track the changes from one 2D image to another; usually from adjacent frames in a video sequence. This technique can be used to piece together the events of a car accident or a home invasion. 7

8 IMAGE RESTORATION takes a corrupt/noisy image and transforms it back to the clean, original image. This technique is performed by imaging a point source and using the Point Spread Function (PSF) to restore the image information lost to the blurring process. Image restoration is especially useful when images must be clarified for optimized processing. 8

9 Augmented Reality s Link to Computer Vision Via the use of computer vision, all consumer-facing interactions can be significantly enhanced with the introduction of Augmented Reality (AR). Through mobile technology, AR the ability to overlay and share physical objects, spaces and images on mobile devices enables customers to interact three-dimensionally with physical objects or their surroundings. Predicted by Goldman Sachs to be an $80B market by 2025, AR serves as a powerful new mechanism for insurance companies to service their customers. For example, AR can be used to visually guide customers to install new smart security systems, provide customers with line-by-line explanations about their contract, or walk customers through the process of submitting a claim in self-service mode. 9

10 Computer Vision s Potential in P&C Insurance The insurance industry is poised to benefit from the advantages inherent in computer vision. Visuals have always been an important component of processing a claim. In most cases, a representative from the insurance company agent, assessor or adjuster is required to visit a site to collect visual evidence to solidify their decisions. Unlike other industries, this visual evidence is stored, resulting in insurance companies amassing vast visual resources of images of insured assets and subsequent damage. Another factor is the industry s transition towards digital self-service. Recognizing the efficiencies and cost saving inherent in self-service, 49% of insurers have already integrated digital claims processing systems with first notice of loss (FNOL) systems. The shift to digital claims has been proven to increase efficiency; according to McKinsey, a digital claims function can drive a 20% increase in customer satisfaction score and 25-30% reduction in claim expenses. With digital technology in place, the insurance industry has successfully laid the foundation for future opportunities that incorporate the semi-automated and fully automated claims processes that are part of the computer vision framework. 10

11 Computer Vision in Customer Service Customer Service is already shifting toward visual communication thanks to the popularity of video chat technologies and video tutorials. Adding the element of computer vision AI will take customer service to the next level with automation. Computer vision can add essential data to customer s profile based on visual data: the customer s facial profile, home environment, car, purchased devices, bills, contracts, etc. It can also help predict risk and assess damage, allowing the insurance company to make customer-facing decisions more quickly and accurately. Computer vision technology can be implemented by insurance companies in assisted service mode to route visual customer inquiries, interpret them and assist the agent or adjuster with visual decision support tools or in self-service mode where the customers visually interact with virtual assistants that are able to visually guide them through the claims process. 11

12 Assisted Service Computer vision can be utilized to perform as a virtual assistant for customer service agents, delivering effective decision support during the agent-customer interaction. Considered a hybrid model, the agent s performance is enhanced by the computer s ability to quickly identify objects and assess damage. This model is especially effective when the insurance company is required to handle large call volumes of repetitive inquiries as the efficient process leads to shorter handling times, higher FNOL resolutions and a more satisfying customer experience. With assisted service, the agent accesses the required visuals either through a live visual session with the customer via his mobile screen, or in offline mode where the customer uploads images via the app/ website or provides images during interaction with IVR while waiting for the live agent. Computer vision technology can then analyze the images and provide effective decision support to the agent. 12

13 Self-Service Integrated within a company s self-service apps or website, the goal of the visual assistant is to enable full self-service with object recognition and augmentation. Via a smartphone, the customer indicates the object or environment in question, and the visual assistant can recognize the object and provide automated guidance. Augmented reality is used to guide the customer via a step-by-step process and is also able to correct the customer in case of errors, ensuring that the process is successful. 13

14 The Synergy between Assisted and Self Service There is a tight link between the two modes of service. In cases where the visual assistant has failed to properly analyze the video/image, a seamless handover to agents/adjusters can be executed. In these solutions, the visual assistant learns from the agent s inputs and improves the automated responses over time. Examples of the Synergy For example, take the case of damage to a vehicle following a car accident. The customer opens a link on the insurer s website to file a claim. If the damage is light, the driver can easily utilize a self-service app that will guide him through the process of uploading images of the damage and the basic documentation process. He will be instructed to take multiple photos of the vehicle, including the license plate and photos from specific angles. All images are automatically stamped with geo-location and time. The app will then use facial recognition technology to establish positive identification of the driver, reducing the risk of fraud. The system can also isolate new damage from old, existing damage by cross-referencing historical images stored in its database. If the damage is more severe, all driver-provided images will be saved and transitioned to a live agent. The agent receives the data from the self-service app including all auto-analysis of the images and damage, authenticated driver identification via facial recognition, and a report indicated suggested compensation based on the damage analysis. The agent can then complete the claims process. 14

15 Potential Insurance Use Cases Incorporating computer vision into the claims process significantly enhances the way insurance-related inquiries are processed and resolved supporting the entire claims cycle across a wide variety of uses cases. DAMAGE RECOGNITION Assisted service Self-service Following a car accident, a live video connection with the customer can be established in order for the customer to show an insurance agent the cause and extent of the damage. Computer vision recognizes and analyzes the damage to the vehicle and utilizes its AI capabilities to estimate the cost of the damage based on historical reference. Customers can upload images and videos of the damage for rapid, accurate assessment. For example, if a windshield was cracked as a result of a car accident, the computer visionbased visual assistant can detect the damage and guide the customer through the process of capturing the required images. Once the visual evidence is uploaded, computer vision can assess the cost of the payout and validate the claim. ONBOARDING/UNDERWRITING Assisted service Self-service A new customer would like to open an insurance policy for his baby grand piano. Without computer vision, insurance companies would have to perform expensive surveys or issue risky policies. With computer vision technology, adjusters can verify and document the insured asset s status when accepting the customer and avoid the need to send estimators, effectively reducing risk and costs. A customer would like to onboard a car to his existing insurance policy. The visual assistant can guide him through the process of capturing images of the car from all angles by recognizing car parts, while detecting existing damage. This process enables the insurance company to remotely verify and underwrite a new asset, storing the images for future reference. 15

16 BILLING AND CONTRACTING Assisted service Self-service A customer would like to understand specific line items on the 6-page bill he just received from the hospital following an extensive stay. With computer vision, the visual assistant detects the highlighted sections, saving the agent time in guiding the client through the paperwork. A customer wants to make some changes to his billing preferences, and ID authentication is required for him to do so. With computer vision, the visual assistant guides the customer through the form required for the identification process. The customer uploads a selfie, and facial recognition technology positively validates his identity, allowing him to proceed with his account updates. INSURANCE IOT Assisted service Self-service A customer needs help troubleshooting his new smart security camera. A customer complains that his smart security camera has ceased functioning. The agent captures the image of the security camera, enabling computer vision to identify the device and its technical issue, and then support the agent in suggesting next-step instructions for a resolution. A customer needs help installing a new smart security camera. Using computer vision, the insurance company can see the unit and any associated error message, analyze the issue and guide the customer to the resolution using AR. CLAIMS ADJUSTMENT Assisted service Self-service A customer contacts his insurance agent to open a claim for flood damage to his apartment. Adjusters can see the damage via the homeowner s smartphone, and computer vision can analyze and assess the extent of the water damage, enabling the adjusters to make decisions from their office, and eliminating the need for travel time and expenses. A winter storm caused a tree to damage a customer s deck. The customer can upload images and videos of the damaged deck, computer vision can detect the source of the damage and estimate its cost to repair, enabling the insurance company to quickly process of the claim. 16

17 The Value Computer Vision Brings to Insurers Reduced claim settlement cycle time: Computer vision s ability to validate claims in real-time enables expedited claim settlements often after the First Notice of Loss (FNOL). Lowered cost per claim: Computer vision improves KPIs across the board, which translates into significant cost reductions in contact center from improved FCR and AHT, reduced adjuster time in the field, and less customer churn More accurate appraisals: When computer vision can reference a robust knowledge bank filled with masses of classified historical images, insurance companies have the upper hand in delivering accurate appraisals and determining insurance premiums. Reduced adjuster travel time and cost: With the ability to connect visually with the customer, and analyze important visuals remotely, agents can process claims from the comfort of their office without dispatching adjusters to the field. 17

18 Increased number of claims processed per employee: With a faster AHT and increased FCR, insurance companies can validate and process claims faster than ever without the need for staff augmentation. Less fraudulent claims: The use of facial and image recognition technologies, as well as images stamped with geo-location and time, enables insurance companies to reduce their risk of fraud. Enhanced customer satisfaction: The visual upload process is fast, interactive, and eliminates the need to wait for a field adjustor or live agent to approve the claim. When claims are processed faster, customers are happier. Easy adoption of insurance smart devices: Installing smart devices in a self-service mode is the key to their adoption. Using computer vision, the insurance company can see the device, recognize its parts and any associated error messages, and guide the customer through the installation, setup or troubleshooting process. 18

19 The Challenges of Implementation Computer vision is still at a very early stage, and far from being widely adopted. While Gartner estimates that the use of AI in the insurance industry is low at about 10% there are pioneering companies who are successfully using AI-based video analysis to improve their business operations. For more widespread adoption, Gartner believes that the innovation must be led by the business and not IT in order to overcome the natural resistance to change. The challenges do not end with implementation. For machines to recognize and analyze visual images to the highest degree of accuracy, the creation of massive data sets is required to effectively train the model. For example, in order for the visual assistant to correctly identify damage, the machine must have had the opportunity to process tens of thousands of images of each vehicle model in various lighting, angles and positions. While some companies store their visual data in several use cases, for some others, building these massive data sets is extremely time consuming and laborintensive, and simply out of scope. In addition, privacy is an issue high on public consciousness. Visuals acquired from a customer s home or property trigger a number of privacy and security issues. For example, images uploaded by the customer or captured by the insurer may include background images of family members, the home s physical environment, or other identifiable personal details. The enterprise may be required to anonymize images or delete images with identifying elements such as customer s faces. Due to the efforts required for data acquisition and these privacy issues, it will take time to select the right use cases, dedicate budgets and move from a POC stage to a full production. 19

20 How to get started implementing Computer Vision in insurance BRING THE BUSINESS ON BOARD While tech-focused, computer vision projects should not be driven by IT alone. The business whether C-level execs, the Board or CX management should invest the necessary time and resources to understand these technologies and the value they bring to the company. CREATE AND IMPLEMENT A DATA STRATEGY As the most valuable organizational asset, data acquired during a policy s life cycle is how insurers identify, analyze and manage risk. Ensure that the company has sufficient data, and the quality of that data is sufficient and available for use without privacy concerns, for example. Insurance companies must develop a clear and actionable strategy for obtaining and securing access to external data, as well as ways to combine this data with internal sources. IDENTIFY USE CASES WHERE COMPUTER VISION CAN ADD VALUE Senior leadership must then determine how computer vision can best support their longterm strategic roadmap. It is vitally important to select and define a use case or POC that will both demonstrate value and ROI. Use cases should be determined based on business need, availability of visuals, complexity, volume of calls and other factors. For example, an insurance company must decide if computer vision can bring more value to vehicle-related claims or home-related claims. DETERMINE WHICH PLATFORM FEATURES ARE NECESSARY On the most basic level, the platform must efficiently connect customer service agents and the customer with minimal friction. A robust knowledge-base should feature masses of classified and tagged images, a repository of industry-based data, a visual agent reference library for best-path decisions, embedded visual call script guidance, and an agent training application. Robust security and privacy protocols and agent management features should be available as well. 20

21 EXPLORE WHAT RELEVANT SOLUTIONS ARE AVAILABLE FOR YOUR TYPE OF NEED End-to-end platforms can be delivered and implemented by external suppliers; these vary greatly by cost and capabilities. Explore your options by scheduling demos, often available for free, as well as POCs and pilot deployments. Always do your due diligence and carefully check customer references. HIRE THE RIGHT TALENT Make sure you have staff with the right combination of skill set and mindset to drive computer vision projects forward. Whether outsourced, hired or trained, these roles will include data engineers, data scientists, technologists, cloud computing specialists and CX designers who work together as a team to deliver holistic customer experiences. Efforts will need to be expended on culture shift and change management to facilitate this evolution. 21

22 Summary Claims processing is a key touch point in the customer life cycle, and insurers have only a short opportunity to wow or disappoint their policyholders. With much at stake, insurers must adopt innovative tools and technologies in order to transform the claims function. Implementing computer vision technology that enables visual insurance claims has emerged as a solution that can transform the entire claims process. Computer vision guides customers through the process of capturing visuals while recognizing objects within the images, classifying the images, and routing the inquiries to the to the right agents or adjusters for immediate auto-analysis and incident assessment. The video element is integrated into the existing workflow, in sync with existing systems, resulting in a significantly shorter and more straightforward claims process that delivers real value to P&L insurers. For insurance companies looking for an advantage in today s highly competitive insurance market, implementing computer vision-based claims will not only boost the customer experience, but their ROI as well. 22

23 ABOUT TECHSEE TechSee revolutionizes the customer support domain by providing the first cognitive visual support solution powered by augmented reality and artificial intelligence. TechSee empowers support teams across the globe to deliver a visual customer experience that significantly reduces service costs, enhances service quality, and delivers intelligent fully-automated services over time. TechSee is led by customer service industry veterans with years of experience in customer experience technologies, computer vision and big data. Techsee is headquartered in Tel Aviv with offices in Boston and Madrid..