Legal

    Billing Automation for a Legal Firm

    We built an autonomous agent that processes invoices, extracts key data, and updates the ERP without human intervention, reducing handling time by 80%.

    Back to case studies
    Automation
    LLM
    OCR
    ERP

    Bufete Mediterráneo

    November 2025

    Challenge

    The firm was processing more than 500 invoices per month manually, with a three-person admin team spending four hours a day on the task. Data entry errors were creating discrepancies in financial reports.

    Solution

    We implemented a multi-agent pipeline with OCR plus LLM extraction, a consistency validator agent, and a direct ERP connector through a REST API. The system processes invoices 24/7, with human supervision only for low-confidence cases.

    Results

    80%
    Time reduction
    Fewer hours spent on manual invoicing
    99.2%
    Accuracy
    Correct data extraction rate
    340%
    Year-one ROI
    Return on investment over 12 months
    2,400
    Invoices/month
    Maximum processing capacity

    Architecture

    1

    Ingestion

    Invoices received by email or shared folder

    Python
    IMAP
    Watchdog
    2

    OCR extraction

    Convert PDF or image files into structured text

    Tesseract
    PDFMiner
    OpenCV
    3

    LLM agent

    Semantic field extraction with GPT-4o

    OpenAI API
    LangChain
    Pydantic
    4

    Validation

    Consistency checks and business-rule verification

    Python
    Custom Rules Engine
    5

    ERP integration

    Automatic handoff into the accounting system

    REST API
    OAuth2
    Holded

    Want to build something similar?

    If this case sounds familiar, we can help you define a version adapted to your business.

    Talk about my case