M. Stirm II
Case Study · AI Workflow Engineering

AI-Assisted
Legal Document
Workflow

A modular AI-powered workflow that automates legal document intake, OCR, document intelligence, translation, glossary generation, review package creation, and workflow orchestration — while preserving human oversight.

Explore FeaturesWatch Demo
n8nClaudeAnthropicGoogle DriveGoogle DocsGoogle SheetsLovableOCRAI AutomationPrompt Engineering
Business Problem

The Challenge

Legal translation providers spend significant time on repetitive administrative work before any expert linguistic effort begins.

  • Intake
    Manual routing of new client submissions.
  • File Organization
    Ad-hoc folder structures per matter.
  • Language Detection
    Human triage of unknown source languages.
  • Document Classification
    Sorting contracts, filings and correspondence by hand.
  • OCR
    Running scans through separate tools.
  • Glossary Creation
    Rebuilding terminology per project.
  • Review Packet Assembly
    Bundling artefacts by hand for reviewers.
  • Job Tracking
    Status updates scattered across email and spreadsheets.
Key Features

What the Workflow Ships With

AI OCR

High-accuracy text extraction across scans and photos.

Automatic Language Detection

Identifies source languages before routing.

Document Classification

Sorts contracts, filings, correspondence and more.

Deterministic PII Masking

Reproducible redaction using rule-based masking.

AI-assisted PII Review

Model-flagged entities queued for human confirmation.

Legal Translation

Domain-tuned prompts preserve legal terminology.

Glossary Generation

Per-project term banks extracted automatically.

Google Drive Integration

Native folder structure for every matter.

Review Packet Generation

Bundled artifacts ready for human sign-off.

PDF Export

Deliverables formatted for downstream systems.

Translation Dashboard

Live status across every job in flight.

Lovable Client Portal

Branded intake surface for end clients.

Engineering Decisions

Design Decisions

The trade-offs behind the architecture.

Each stage is an independently deployable node. Swap an OCR provider, upgrade a translation prompt, or add a new classifier without touching the rest of the pipeline.
Business Value

Qualitative Outcomes

Impact framed by capability, not vanity metrics.

Reduce repetitive administrative work

Improve workflow consistency

Increase scalability

Maintain human quality control

Accelerate document preparation

Improve terminology consistency

Technical Skills

Skills Demonstrated

AI Engineering
Prompt EngineeringClaudeLLMsOCRDocument Intelligence
Workflow Automation
n8nGoogle Workspace APIsAutomation DesignAPI Integration
Software Design
Modular ArchitectureError HandlingHuman-Centered AI
System Design
Privacy EngineeringPipeline ReliabilityScalable Components
Project Gallery

Screenshots

Interface captures across the workflow.

Lovable Portal
Lovable Portal
n8n Workflow
n8n Workflow
Translation Dashboard
Translation Dashboard
Glossary Output
Glossary Output
Generated Review Package
Generated Review Package
Video Walkthrough

Watch the Complete Project Walkthrough

A complete walkthrough of the architecture, workflow, engineering decisions, and live demonstration.

About This Project

The intent behind the MVP

This project was developed as an MVP demonstrating reusable AI workflow architecture rather than a production deployment. It prioritises modular design, privacy-first engineering, human-centered AI, scalable automation, and maintainable architecture — the qualities that make an AI system trustworthy inside a real professional workflow.

Let’s Talk

Interested in AI Workflow Automation?

I'm passionate about designing human-centered AI workflows that automate repetitive work while preserving expert oversight. If you'd like to discuss this project or explore opportunities to collaborate, I'd love to connect.

View Portfolio Contact Me