Inspiration
Many SMEs and startups find it difficult to stay on top of the ever-changing environmental, social, and governance (ESG) regulations and compliance standards. Legal advice is costly and time-consuming, and creating policies by hand is prone to mistakes. In order to minimize human labor and increase accuracy, we aimed to develop an autonomous AI agent that could intelligently generate, verify, and maintain compliance policies in real-time.
What it does
Auto Policy AI is a self-governing compliance tool that: Takes as inputs the location, industry, and company profile. Uses web scraping and APIs to retrieve pertinent regulations and ESG guidelines. Prioritizes tasks and finds gaps and conflicts using LLM reasoning. Produces policy documents in a variety of formats that are actionable and audit-ready. Proposes remedial actions for identified gaps in compliance. Updates policies frequently in response to changing regulations.
How we built it
Reasoning & LLM Hosting: Sage Maker AI & Amazon Bedrock Agent Autonomy: Using Amazon Bedrock Agent Core primitives to assess ESG risks and create policy proposals. Serverless execution: tasks are scheduled and updates are retrieved using AWS Lambda functions. Storage: Created policies and old logs are kept on Amazon S3. API Interface: Agent as a Service is exposed via the Amazon API Gateway. Data processing: NumPy, Pandas, and Python for managing structured data. External Integration: Web scraping and regulatory APIs for up-to-date legal information.
Challenges we ran into
Managing multi-source, dynamic regulations and converting them into requirements for practical policy. Ensuring that conflicts between overlapping laws and standards could be automatically resolved by the LLM reasoning pipeline. Testing agent autonomy while controlling endpoint latency and AWS service quotas. Creating an architecture that is scalable to accommodate various sectors and legal systems.
Accomplishments that we're proud of
Created a fully functional autonomous AI agent that can create compliance and ESG policies. Agent Core and LLM reasoning were successfully combined to enable multi-step decision-making. Created a fully tested and VS Code-coded AWS architecture that is ready for deployment. In a roughly three-minute demonstration, real-time policy creation and conflict resolution were demonstrated.
What we learned
How to use Agent Core, Sage Maker AI, and AWS Bedrock to create autonomous agents. Methods for using LLM-driven logic to resolve challenging regulatory issues. AWS serverless architecture best practices for autonomous real-time systems. The significance of feedback loops and ongoing learning in the creation of autonomous policies.
What's next for Auto Policy AI – ESG & Compliance Agent
Include multilingual support for international laws. Allow the agent to automatically manage risk assessments unique to a given industry. Include dashboards for real-time compliance breach monitoring and alerting. Examine research publication opportunities on self-governing AI agents for automating corporate compliance and creating ESG policies.
Log in or sign up for Devpost to join the conversation.