Fast Tracks to Compensation: Revolutionising Rail Claims with Drupal and AI - A Case Study
In a world where time is money, discover how we revolutionised the rail industry's Delay Repay claims management system using Drupal and AI, and reduced their processing time from 30 minutes to 5. Join us on a journey of innovation, efficiency, and the transformative power of AI in business applications.
Prerequisite
Only a general interest in AI is required, but attendees will benefit more if they have an interest or knowledge in business applications and claims management systems.
Outline
In the UK alone, there were 8 million claims for train delay compensation in 2023. When our company FreelyGive first entered the market in 2018, a Which? Investigation showed that delay repayments were badly mismanaged and most train companies had long backlogs of claims with passengers not getting their compensation.
We were able to use native Drupal to build a complete delay compensation system, which handled the webforms, a database of train movements, repayment, fraud detection and reporting all using Drupal. Provided directly to the Train Companies, this brought the average processing time for each claim down from 30 minutes to only five.
This presentation will focus on how we were able to improve things further with AI and sell the concept to the train companies. We discovered that the initial automation was easy but proving the safety of AI to our clients was more complicated and required clear explanations of prompt engineering and evaluations based on past data. We were able to achieve accuracy ratings similar to humans and reach automation rates of 70% - 90%.
I will present a case study outlining the specific unique areas of a delay repay system and how we used Drupal to build it. We will then show the tools, modules and reports we used to automate claims using AI and outline the process we went through to improve the prompts by applying machine learning to real data (i.e. thousands of previous claims processed by humans into the system). Whilst the Delay Repay system is proprietary, the core modules used to build it are open source and released on drupal.org.
The goal of this case study is to help individuals see how they can use Drupal in a novel way to automate processes across multiple industries.
Learning Objectives
- Understanding how Drupal can be used to manage a claims system.
- Understanding the specific Drupal modules that can be used to create LLM driven workflows.
- Understanding the importance of Evals (Evaluations) in building AI systems and sell them to management.
Experience level
Beginner
Prerequisite
Only a general interest in AI is required, but attendees will benefit more if they have an interest or knowledge in business applications and claims management systems.
Outline
In the UK alone, there were 8 million claims for train delay compensation in 2023. When our company FreelyGive first entered the market in 2018, a Which? Investigation showed that delay repayments were badly mismanaged and most train companies had long backlogs of claims with passengers not getting their compensation.
We were able to use native Drupal to build a complete delay compensation system, which handled the webforms, a database of train movements, repayment, fraud detection and reporting all using Drupal. Provided directly to the Train Companies, this brought the average processing time for each claim down from 30 minutes to only five.
This presentation will focus on how we were able to improve things further with AI and sell the concept to the train companies. We discovered that the initial automation was easy but proving the safety of AI to our clients was more complicated and required clear explanations of prompt engineering and evaluations based on past data. We were able to achieve accuracy ratings similar to humans and reach automation rates of 70% - 90%.
I will present a case study outlining the specific unique areas of a delay repay system and how we used Drupal to build it. We will then show the tools, modules and reports we used to automate claims using AI and outline the process we went through to improve the prompts by applying machine learning to real data (i.e. thousands of previous claims processed by humans into the system). Whilst the Delay Repay system is proprietary, the core modules used to build it are open source and released on drupal.org.
The goal of this case study is to help individuals see how they can use Drupal in a novel way to automate processes across multiple industries.
Learning Objectives
- Understanding how Drupal can be used to manage a claims system.
- Understanding the specific Drupal modules that can be used to create LLM driven workflows.
- Understanding the importance of Evals (Evaluations) in building AI systems and sell them to management.
Experience level
Beginner