Arun

Home

Home

Work

Work

About Me

About Me

Contact

Contact

Download Resume

Arun

Home

Home

Work

Work

About Me

About Me

Contact

Contact

Download Resume

Arun

Work

Work

Download Resume

Cognitive Claim Assistant (CCA)

Computer Vision based vehicle image analysis, Machine Learning models & Robotic Process Automation (RPA) to automatically produce a detailed damage cost estimate in real time

My Role

Individual contributor.

As a User Experience designer, I teamed up with Business stakeholders, the Product Manager, Product Owner, Engineering Team, QA Team, and Release Team.

To start with we had a product workshop to understand the scope, create ideas, and to know more about the technology capabilities. Involved in presenting ideas to the stakeholders, User Research, Interaction Design, Visual Design, collaboration with the engineering team to have a final product per the approved designs, and Usability Testing.

Domain/Industry

Auto Claims & Auto Finance

Device/Platform

Web Application

Tools Used

Sketch, Invision, Figma,
Zeplin, Azure, Miro, Zoom

Duration

9 Months,
Kick-off -> MVP -> Release v1.0

What problem did we solve?

Insurance industry is being disrupted due to a changing landscape and Digital is one of the key forces challenging the current operating model

In today's highly competitive insurance industry, efficient and accurate claims processing is crucial for customer satisfaction and retentionA

  • Loss of revenues and market share as customer retention impacted by long drawn process, multiple touch points and poor customer experience.

  • Estimate leakage due to non standardised & inaccurate cost estimates dependent on human judgment and inability to detect fraudulent claims upfront in the process.

  • Huge estimation costs due to the time put in by costly resources (adjusters) to prepare a damage cost estimate.

Why we want to solve this problem?

Leading to disruptive business outcomes and a transformative experience for the end customer.

Overall Cycle time reduced

bringing down the rental,

storage, towing and

adjustor costs

Overall Cycle time reduced bringing down the rental, storage, towing and adjustor costs

Accurate estimation of

repair costs through AI

based holistic analysis of

multiple data sources

Accurate estimation of repair costs through AI based holistic analysis of multiple data sources

Loss Ratio improved

substantially as potential

leakage in estimate sheet

is removed

Loss Ratio improved substantially as potential leakage in estimate sheet is removed

Fraudulent Claims can be

detected upfront through

past lifecycle data thus

saving costs

Fraudulent Claims can be detected upfront through past lifecycle data thus saving costs

Market differentiation

leading to revenue increase

and customer retention


Market differentiation leading to revenue increase and customer retention

My Design Process

As a team we have defined the business goals in the form of problem statement. I planned to conduct research to gain a deep understanding of the users and their needs. This helped me to know the users and understand what they need from the product.

Research

Analysis

Ideation

Design

Feedback

Iteration +

Handoff

Stakeholder Interviews

User Interviews

Competitive Audit

Persona’s

User Needs & Goals

Problem Statement

Brainstorm

Feature definition

User Flow

Sitemap

Sketches

Wireframe

Visual Design

Prototype

Design Critique

Business Demo

Usability Testing

Iterations

Engineering Demo

Assets Hand off

UX Audit

Feedback

Iteration +

Handoff

Design Critique

Business Demo

Usability Testing

Iterations

Engineering Demo

Assets Hand off

UX Audit

My Design Process: Research

My UX Research plan was to talk to stakeholders to understand the business goals, then proceeded with the user interviews to know about their needs, goals and pain points. Followed by the competitor analysis to study how users interact with the existing application.

Stakeholders (Partners, PM, PO, SME), are the first level of information providers. They help in validating the hypothesis that keeps in align with the business goals.

Inaccurate manual classification of claim cases

Inaccurate manual classification of claim cases

Poor customer experience

Poor customer experience

Potential leakage in estimates

Potential leakage in estimates

Body repair shops not held accountable

Body repair shops not held accountable

Dependence on external appraisers or repair shops

Dependence on external appraisers or repair shops

Long time taken

Long time taken

Inaccurate estimates

Inaccurate estimates

Lack of transparency

Lack of transparency

User Interviews - In total i got to speak with 16 users (Appraisers, Reviewer’s & in-house SME). Interviews helped to valid the hypothesis and validating of the business goals.

Supplements are a real pain as they come back late

Supplements are a real pain as they come back late

Independent appraisers: 3 day turnaround time.

Independent appraisers: 3 day turnaround time.

No patterns of receiving the claims (come in bulk or like emails)

No patterns of receiving the claims (come in bulk or

like emails)

Locating parts + Part Availability is also a pain.

Locating parts + Part Availability is also a pain.

10 to 11 days per claim

10 to 11 days per claim

Competitive Audit, is performed on the following application.

  • Brightclaim.

  • Claimgenius.

Clutter UI

Hard to follow the flow and progress

Hard to follow the flow and progress

No alerts functions

Confusing Nav

Lack of visual consistency

Out dated design

My Design Process: Analysis

After the research process I was able to outline persona's (Desk appraiser and Reviewer)

Business Goals;


Team wants a solution that helps the business and users to provide application in helping with the claim process.


  • Damaged part (E.g. Bumper, left door)

  • Likelihood if damaged or not

  • Severity of damage (Repair or Replace)

  • Total Loss (Yes/No)

  • Cost Estimate for the repair of vehicle

  • Fraud probability

Business Goals;


Team wants a solution that helps the business and users to provide application in helping with the claim process.


  • Damaged part (E.g. Bumper, left door)

  • Likelihood if damaged or not

  • Severity of damage (Repair or Replace)

  • Total Loss (Yes/No)

  • Cost Estimate for the repair of vehicle

  • Fraud probability

** Synthesize

With all the notes and recordings from the interviews, I put everything in perspective, synthesized with

the participants, and created the personas, captured users needs, goals and pain points

My Design Process: Ideation

Solution flow high level

First Notice of

Loss (FNOL)

First Notice of

Loss (FNOL)

First

Notice of

Loss (FNOL)

Images of damaged

vehicle

Images of damaged

vehicle

Images of

damaged

vehicle

CCA

application

CCA

application

CCA

App

Repairable

Repairable

Total Loss

Total Loss

Parts cost options (for all damaged parts)

Predict Total cost estimate value

Predict Repair Cycle Time Rental,

Storage Costs

Labor rates data

Past repair history data

Parts cost options (for all damaged parts)

Predict Total cost estimate value

Predict Repair Cycle Time Rental,

Storage Costs

Labor rates data

Past repair history data

Sitemap

User flow

** Brainstorm
I started with getting few ideas after synthesizing the research, that would attempt to solve users’ problems.
This is done with the collaboration with other teams just to get few new ideas and suggestions.

My Design Process: Design & Feedback

Sketch

Low-Fidelity Wireframe

Visual Design

Clickable Prototype

Low-Fidelity Wireframe

**

To put the ideas as a visual representation its always good to start with Pen and paper to sketch and exploring different approaches.


Created 4-5 ideas which is taken to the business for the approvals.

**

To get approval and to start with the process I started creating low-fidelity wireframes based on my sketches. Lo-fi wireframes are quick and easy to do. If there is any comments or suggestions it was easy to make changes and adjustments super fast .

**

After business approval on wireframe,

i started with the visual design following existing design system to make the design consistency across the business offerings.

**

Its time to bring the designs demo able for larger audience and do usability testing with the users. I used invision app to create prototype, then moved everything to Figma.

Visual Design

**

To put the ideas as a visual representation its always good to start with Pen and paper to sketch and exploring different approaches.


Created 4-5 ideas which is taken to the business for the approvals.

Clickable Prototype

**

To get approval and to start with the process I started creating low-fidelity wireframes based on my sketches. Lo-fi wireframes are quick and easy to do. If there is any comments or suggestions it was easy to make changes and adjustments super fast .

**

After business approval on wireframe,

i started with the visual design following existing design system to make the design consistency across the business offerings.

Concept Sketch's & Wireframe

Here are few concept sketchs presented to the business.

Visual Designs

The final designs for one of the approved solutions.

Solution:

Detailed view of the claims that have been assigned to desk appraiser after Machine Learning based Computer Vision technology analyzed images with other details.

Solution:

Detailed view of the claims that have been assigned to desk appraiser after Machine Learning based Computer Vision technology analyzed images with other details.

Impact:

  • Intelligent classification of total loss cases upfront can reduce claims cycle time substantially.

  • Potential $100-$300/claim savings for every claim case which is prevented from getting re-routed.

  • Better customer experience leading to customer retention and more market share.

Impact:

  • Intelligent classification of total loss cases upfront can reduce claims cycle time substantially.

  • Potential $100-$300/claim savings for every claim case which is prevented from getting re-routed.

  • Better customer experience leading to customer retention and more market share.

Solution:

With the AI analyzied images it predict whether the claim is either Repairable or Total Loss.

Solution:

With the AI analyzied images it predict whether the claim is either Repairable or Total Loss.

Impact:

  • Decrease the time taken to write an estimate by ~50%

  • Reduce cost of writing an estimate by ~50%

  • Improves the overall cycle time of claims processing leading to cost savings and better customer experience.

Impact:

  • Decrease the time taken to write an estimate by ~50%

  • Reduce cost of writing an estimate by ~50%

  • Improves the overall cycle time of claims processing leading to cost savings and better customer experience.

Solution:

Desk appraiser has the ability to view to the overall damages for the particular claim in the detail vehicle parts view. This has the ability to show all the relevant information.

Solution:

Desk appraiser has the ability to view to the overall damages for the particular claim in the detail vehicle parts view. This has the ability to show all the relevant information.

Impact:

  • This reduces the cost and time for an field appraiser or eliminate of their service to visit the place where the car is kept.

Impact:

  • This reduces the cost and time for an field appraiser or eliminate of their service to visit the place where the car is kept.

Solution:

Display all the damaged parts with the damage type, % of damage and decision for the particular part. This helps the appraiser because the manual estimates are prone to human errors in either repair time calculation.

Solution:

Display all the damaged parts with the damage type, % of damage and decision for the particular part. This helps the appraiser because the manual estimates are prone to human errors in either repair time calculation.

Impact:

  • Appraiser can get the replaceable parts details and its availability.

  • Accountability across repair shops/appraisers by providing gross estimate and repair time baseline upfront

Impact:

  • Appraiser can get the replaceable parts details and its availability.

  • Accountability across repair shops/appraisers by providing gross estimate and repair time baseline upfront

Solution:

DeskAppraiser have an ability to change the AI predicted output and its values incase appraiser this the prediction is incorrect.

Solution:

DeskAppraiser have an ability to change the AI predicted output and its values incase appraiser this the prediction is incorrect.

Impact:

  • Desk appraiser can check the price quote from different 3rd party vendors.

  • This reduce the cost on body shop rent costs for the customers

Impact:

  • Desk appraiser can check the price quote from different 3rd party vendors.

  • This reduce the cost on body shop rent costs for the customers

Solution:

DeskAppraiser have an ability to change the AI predicted output and its values incase appraiser this the prediction is incorrect.

Solution:

DeskAppraiser have an ability to change the AI predicted output and its values incase appraiser this the prediction is incorrect.

Impact:

  • Increase in profitability (lower Loss ratio) as damage estimate generated is highly accurate.

  • Cycle time for processing claims can dramatically go down through this solution.

  • Ease of scaling up without the need of waiting to hire and train new human appraisers

Impact:

  • Increase in profitability (lower Loss ratio) as damage estimate generated is highly accurate.

  • Cycle time for processing claims can dramatically go down through this solution.

  • Ease of scaling up without the need of waiting to hire and train new human appraisers

Outcomes & takeaway

Outcomes & Takeaway

Business & User Satisfaction:

Business & User

Satisfaction:

Partner & Business team is happy about the outcome of the product. This is highly taken by the higher levels in the company.


Users are very much interested to see this product in use soon.

AI Prediction Accuracy

AI prediction percentage is around 76% which little lower than the industry standards. Data analytics team is focusing on improving the overall percentage by labelling more number of images from the repository.


Team, tested with the vehicles in Indian roads too.

New Engagement

+3

This solution has given business an opportunity to bring new clients/customers to utilise the product capabilities for their company in auto mobile sectors.


* One of the US leading car rental service company have tested to check their damages after the car is returned.

Learnings:

  • I got an opportunity to work and understand how machine language works and its impact to the application through technology.

  • Collaboration is also important aspect in User Research.

  • This is being my first product company i got an experience how an product evolve from initial stage till the release of the product live.

Challanges:

  • I need to work on many iterations which is good in a way. we were able to bring the desired output for both users and business.

  • I did see few technical constraints from the engineering team when the design is showcased to them for the development.

  • Getting users time and availability for the user research is tedious task.

Appreciation

Our team and the solution of the product is highlighted in one of the online magazine.

... and also accolades from the company for our team contribution.

Thank You!

© arunprakash - 2023 | Created in Framer by "Self Learning" | #Smilemore #BeHappy #Respect

© arunprakash - 2023 | Created in Framer by "Self Learning" | #Smilemore #BeHappy #Respect

© arunprakash - 2023 | Created in Framer by "Self Learning" | #Smilemore #BeHappy #Respect