case study ey vco ai header

EY VCO AI Chatbot

Utilizing AI to serve as a Virtual Compliance Officer for an Enterprise

company_ey_white

Expertise

AI Development

Platforms

Web

Industry

Compliance

The Client

EY - one of the world's largest professional services networks, offers a wide range of services including consulting, assurance, tax, and transactional assistance. As a global corporation, EY cooperates with multinational clients and must navigate and adhere to diverse local regulations globally. 

The Challenge

According to the EU Directive on whistleblower protection, large enterprises are required to introduce channels for reporting violations and mechanisms for whistleblower protection. In 2020, we helped EY build a Virtual Compliance Officer (VCO), a whistleblowing platform enabling anonymous misconduct reporting. Initially, compliance officers were given a variety of tools to smoothly manage all whistleblowing submissions, including a smart documentation repository, intuitive case management, and a chat panel.

Following conversations with their clients, to innovate and improve efficiency of the document repository, EY decided to utilize AI to create its first AI-powered Virtual Compliance Officer. GPT VCO minimizes manual tasks and provides employees with easy access to regulations and AI-powered chat for compliance-related questions.

EL Passion’s AI workflow tool Buildel played a crucial role in integrating the bot within EY's existing systems.

The project blends the newest technology and EY’s specialized expertise and changes the way enterprises approach compliance matters.

AI-Powered Compliance Portal

The project centers around a self-service, AI-powered compliance portal that leverages EY's extensive document repository. The key feature, Resource Augmented Generation (RAG), allows the bot to extract information from the EY VCO database and generate responses based on existing documents. It can also trigger actions such as escalating issues to human officers. 

To ensure the highest response quality, we utilized various AI models. Apart from OpenAI models, this includes an image processing model for extracting text from images and a Multi-language model. One of the biggest challenges in the project was to ensure the bot can generate embeddings, meaning to analyze and compare documents across various languages to verify compliance with both local and international regulations simultaneously. The bot uses GPT-4 for text generation but is structured to easily switch between different models if required.

Integration and Workflow

The bot seamlessly integrates into EY's existing application interface, serving as a Compliance Officer that enhances the chat with AI-driven automation. Our AI workflow tool, BuildEL, serves as the orchestration tool to coordinate user-AI interactions.

Emphasis on Security

To provide data security, the bot operates outside the existing application, maintaining current conversations state, while the VCO stores long-term records. The workflow involves inquiries passing through BuildEL to the bot and back, ensuring a streamlined process.

Tools

  • buildel

    Buildel

  • pgvector

    Pgvector

  • websocket

    WebSocket

  • ruby

    Ruby on Rails

  • elixir

    Elixir

  • phoenix

    Phoenix

  • langchain

    Elixir LangChain

Deliverables

  • Integrating AI into the existing app using Buildel, EL Passion’s AI workflow tool

  • Utilizing Large Language Models GPT-4 and GPT-3.5 Turbo for advanced language understanding and generation

  • Extending the database functionality with effective text-based search with PostgreSQL Vector Extension

  • Image Processing model for text extraction from Images

  • Embeddings generation using MultiLanguage models

  • UX & UI Design: Blending the bot's functionality with human interactions while ensuring consistency with EY's existing visual standards

Maybe it’s the beginning of a beautiful friendship?

We’re available for new projects.

Contact us