AI inspector’app

SAAS, B2B2C, webapp

2023

Lead Designer

PROJECT
Create user experience for B2C

Ravin.AI is a web app that helps customers and professionals check for car damages on a mobile phone with AI and computer analysis.

Launched in April 2021 with Toyota Finance US
2022: KBC Group, Hertz US, Buggy
2023: Rakuten Japan, and Sompo

CHALLENGES

  • The beta app is not intuitive, first-time users are not able to use the app, it requires a training which generates cost.

  • The user are not able to properly capture a scan. The scan generated is poor quality

  • Final reports generated by AI are inaccurate (65% accuracy)
    AI creates false positives during the analysis.

SOLUTION

Created an intuitive, user-friendly experience that enables users and professionals to successfully complete scans, allowing AI to detect damage and generate accurate reports efficiently.

How it works

Result

  • Improved the user experience by 800% (from 1 in 10 users to 9 in 10 users able to complete successfully task)

  • Improved AI accuracy around damage detection by 21%

  • Decreased by 20% manual editing of the report

  • 2022: Named one of TIME’s Best Inventions

Role

Design ownership of the product end-to-end



  • Defined the product strategy, mapped UX, prototyped



  • Ran qualitative user research to gather insights and drove the product prioritization



  • Designed and maintained the design system



  • Supported strategy for key projects through design direction and bringing the vision to life

Design process

Research and observation

Audit in Bulgaria at KBC insurance main office in Sofia.
2 days of observation and analysis of the customer journey from the client and business perspective.

Map user journey

Mapping user flow before translating them into designs helps align the team and take decision on final solution.

Designed and maintained a white label design system

User testing

Ran multiples user research
(interviews and unmoderated test)

Goal : Test user guidance


Dates : June 2021 & August 2021


Participants : 6 customers that had a car with damages


Format : Unmoderated with Usertesting.com


Analysis : Based on the full completion of the task and the feedback of user

Sharing and consolidating learning

Set up a repository to organise user feedback to support the product prioritization

Design features to give more flexibility to professional user
such as editing damages as bulk or manually adding multiple damages

What i learned

• When designing for AI, you need to have a different approach
I highly recommend following Google AI recommendation

• Learnt about AI technical constraint and process

• Discovered the depth and complexity of computer analysis

• Working in a fully remote set-up required to excel in documenting your work and process

• When designing for AI, you need to have a different approach
I highly recommend following Google AI recommendation

• Learnt about AI technical constraint and process

• Discovered the depth and complexity of computer analysis

• Working in a fully remote set-up required to excel in documenting your work and process

• When designing for AI, you need to have a different approach
I highly recommend following Google AI recommendation

• Learnt about AI technical constraint and process

• Discovered the depth and complexity of computer analysis

• Working in a fully remote set-up required to excel in documenting your work and process