DGP Agentic Audit
Project Information
- Category: Compliance Automation
- Client: Pernod Ricard
- Project date: 2025
- Project URL: github.com/albiche/dgp_agentic_audit
Project Overview
The DGP Agentic Audit project delivers a sophisticated, agent-based pipeline to automate compliance audits of Pernod Ricard's digital media presence across over 1000 brand assets. It verifies adherence to 5 mandatory regulatory rules (AAM, FAN, RDM, TRA, UGC) across websites and social media profiles by orchestrating a multi-agent system capable of scraping, analyzing, and scoring compliance in a scalable, robust way.
Key Features & Functionalities
Multi-Platform Extraction
Automated scraping across websites, Instagram, Facebook, X/Twitter, and YouTube, with headless browsing, cookies, and authentication management.
Regulatory Analysis
Detects AAM, FAN, RDM, TRA, and UGC mentions using specialized GPT prompts per platform and content type (text, HTML, images, PDFs).
UGC Link Extraction & Enrichment
Resolves aggregator links (e.g., linktr.ee) and enriches them via GPT to uncover true UGC/Terms URLs and verify compliance policies.
Agent-Orchestrated Pipeline
Coordinates multiple specialized agents for scraping, analysis, OCR, and scoring, with retry and resume capabilities for large-scale audits.
Compliance Scoring
Calculates a regulatory compliance score per URL, indicating the percentage of mandatory mentions detected across formats and platforms.
Configurable & Extensible
All prompts, extraction logic, and rules are data-driven via YAML/JSON, enabling easy adaptation to new regulations or brands.
Architecture & Workflow
The pipeline combines automated scraping (Playwright, API calls) with GPT-4o-based analysis for multi-format content (text, HTML, images, PDFs). Specialized agents handle platform-specific extraction, content parsing, and regulatory analysis. UGC links are enriched, aggregator redirects are resolved, and OCR is applied to images and PDFs. A robust resume system ensures that audits can be restarted without loss. All results are exported in CSV with detailed compliance scores for each URL.
This modular, agentic design allows for easy extension, maintainability, and integration with brand or legal teams’ workflows. All business logic, prompts, and scoring criteria are fully configurable.
Results & Business Impact
The DGP Agentic Audit reduces manual compliance checking costs while ensuring scalable, consistent, and objective regulatory analysis across Pernod Ricard’s entire brand portfolio. By automating detection of required disclaimers and terms, it helps maintain compliance with industry regulations while saving time and resources. The system’s modular, data-driven design also future-proofs it against changing rules and expanding digital channels.