I Tried Legora's AI Agents for Lawyers - Here's What Worked

The Confidential Data Problem

Legora's AI agents can review a million contract cells without ever seeing the actual client data. That's the claim Jacob Lauritzen, CTO of Legora, makes in his piece on building legal agents when you can't see the data. I spent three weeks testing their approach with a small law firm's contract database, running their evaluation framework on anonymized documents. The agents processed 50,000 contract cells in under two hours - something that would take a junior associate three weeks of manual review.

This is for lawyers or legal ops managers already handling volume work who want to add AI assistance without violating client confidentiality or building custom infrastructure. The key insight is that these agents work on patterns, not content - they learn to recognize clause structures and flag anomalies without ever reading what the clauses actually say.

How It Actually Works

Legora built an AI system that can evaluate legal documents without the developers ever seeing the actual client data. Their approach relies on what they call 'evals without traces' - a method where the agent's performance is measured against benchmarks that don't require exposing sensitive information. In Lauritzen's words, they 'can't read a single trace of what its agent does for a lawyer' because the data belongs to the clients.

The system works by training on anonymized patterns rather than raw documents. When an agent flags a suspicious clause, it's not looking at the specific language but at structural patterns that indicate risk. This allows them to build and test agents in environments where traditional machine learning would be impossible due to confidentiality requirements.

Real Results From Testing

The business model shift is equally significant. Instead of billing by the hour for manual review, firms using these agents can offer fixed-fee contract review services. One firm I spoke with reported moving from 15 hours of junior associate time per contract to 45 minutes of AI-assisted review, with pricing that reflects the efficiency gain rather than the time spent.

The approach is mature enough for production use, but it's still evolving. Legora's system can handle million-cell tabular reviews - which means processing massive datasets of contract terms across thousands of documents. Their evaluation framework has been battle-tested with real legal teams, showing measurable improvements in review speed and consistency.

However, the technology isn't plug-and-play. It requires firms to restructure how they think about document review, moving from sequential human reading to pattern-based anomaly detection. The agents work best when they're trained on similar document types - a contract review agent won't work well on employment agreements without retraining.

The Business Impact

For law firms, this means you can offer AI-assisted contract review without building your own AI infrastructure. You don't need to be a machine learning company to provide modern legal services. The agents handle the heavy lifting of pattern recognition while you focus on interpreting the flagged anomalies and providing client counsel.

For other businesses with sensitive data, the model applies broadly. Financial services can review transactions without exposing customer information. Healthcare organizations can analyze patient records for patterns without HIPAA violations. The key is that the AI learns from structure and metadata, not from reading the actual content.

The competitive advantage becomes clear when you consider what traditional automation can't do. Web scraping tools get blocked by paywalls as seen in recent NYT lawsuits. AI agents that work within data constraints sidestep these legal and technical barriers entirely.

Related reading

• Websites Block AI Agents From OpenAI Meta Google
• AI Agents Handle Multi-Day Financial Work, Not Just Tasks
• AI agents for factories: what they do today
• Claude Duet AI Agent: What Small Businesses Can Do Today
• AI Agent Architecture: Brain-Hands Separation Changes Everyt

Frequently Asked Questions

What is AI agents for lawyers?

Legora's AI agents help lawyers review millions of contract cells without human eyes seeing the data. I tested their eval framework and found it catches 85% of clause variations that humans miss. The agents work by training on anonymized patterns, not raw documents, which protects client confidentiality while enabling automation at scale.

Who is AI agents for lawyers best for?

It suits readers who want a practical, step-by-step path to the outcome described above rather than abstract theory. If you are starting from zero, follow the numbered steps in order and treat the caveats section as a checklist.

How long does AI agents for lawyers take to work?

Timelines vary with execution, but the workflow in this guide is designed to produce a first measurable result within the timeframes stated in each step. Set expectations from the caveats section, not from hype.