Studio Founder
How Our AI Agent Reviews Contracts Faster Than Our Lawyers
Key Takeaway
We built an AI agent that reviews contracts in 3 minutes flat β extracting every clause, scoring risk, and benchmarking against market terms β so our lawyers start with answers instead of blank pages.
The Problem
PyratzLabs holds equity in multiple companies. Zama. Artificial-Lab. Others I can't name yet. Each one generates contracts. Vendor agreements, NDAs, licensing deals, partnership terms, employment contracts.
Last year we counted. 140+ contracts crossed our desks. Our external counsel charges β¬350/hour. A thorough contract review takes 4-8 hours. Do the math: β¬1,400-β¬2,800 per contract. Times 140. That's north of β¬250K just to read paper.
The worst part? Most contracts are 80% boilerplate. Lawyers spend hours confirming that standard clauses are, in fact, standard. The 20% that actually matters gets the same attention as the 80% that doesn't.
We needed a filter. Something that could separate the boring from the dangerous before a lawyer ever opens the document.
The Solution
Warren β our CFO agent built on Mr.Chief β routes contracts through the Contract Analyzer skill. Upload a PDF or DOCX. The agent extracts every clause, categorizes it by type, scores risk on a traffic-light system, and benchmarks against market-standard terms.
The agent doesn't replace lawyers. It gives them a 50-page head start compressed into a 2-page brief.
The Process
The workflow is dead simple. Drop a contract into the pipeline:
yamlShow code
# Warren's contract review trigger
skill: contract-analyzer
input:
file: /contracts/incoming/vendor-agreement-acme-2025.pdf
mode: full-review
benchmark: saas-vendor-eu
output:
format: markdown
risk_threshold: yellow # flag yellow and red clauses
The agent runs through four stages:
Stage 1 β Extraction. Every clause pulled, tagged, and categorized:
- Intellectual property assignment/licensing
- Liability and indemnification
- Termination and exit rights
- Payment terms and penalties
- Non-compete and exclusivity
- Data protection and confidentiality
- Governing law and jurisdiction
Stage 2 β Risk Scoring. Each clause gets a color:
View details
π’ Green β Standard market terms. No action needed.
π‘ Yellow β Unusual but not dangerous. Lawyer should review.
π΄ Red β Dangerous. Unfavorable terms, missing protections, or liability traps.
Stage 3 β Benchmarking. The agent compares extracted terms against a database of market-standard positions. A liability cap at 12 months of fees? Green β that's market. Unlimited liability? Red β that's a trap.
Stage 4 β Brief Generation. A structured output the lawyer can scan in 2 minutes:
markdownShow code
## Contract Risk Summary: Acme SaaS Vendor Agreement
**Overall Risk: YELLOW (3 red clauses, 7 yellow, 24 green)**
### π΄ Red Flags
1. Β§7.2 β Unlimited liability for data breaches (market: capped at 12mo fees)
2. Β§12.1 β Auto-renewal with 90-day notice (market: 30 days)
3. Β§15.3 β Governing law: Delaware (your entity is French β enforcement risk)
### π‘ Yellow Flags
1. Β§4.1 β Payment net-60 (market: net-30)
2. Β§9.4 β IP assignment includes derivative works (broad β review scope)
...
The Results
| Metric | Before (Lawyer Only) | After (Agent + Lawyer) |
|---|---|---|
| First-pass review time | 4-8 hours | 3 minutes |
| Cost per review | β¬500-β¬2,000 | β¬200 (lawyer review of brief) |
| Red flags missed | ~5% (fatigue on long contracts) | 0% (systematic extraction) |
| Time to lawyer sign-off | 2-5 business days | Same day |
| Annual review cost (140 contracts) | ~β¬250K | ~β¬35K |
Net savings: ~β¬215K/year. And our lawyers are happier. They review risk briefs, not boilerplate.
Try It Yourself
Install the Contract Analyzer skill on Mr.Chief:
bashShow code
# Install via Mr.Chief dashboard after signing up at mrchief.ai/setup
# clawhub install contract-analyzer
Upload any contract and run:
View details
Review this contract. Extract all clauses, score risk, and benchmark against EU SaaS vendor standards.
Start with a contract you've already had reviewed by a lawyer. Compare the outputs. You'll see.
The agent reads the contract. The lawyer reads the risk brief. Everyone's time goes where it matters.
Related case studies
Studio Founder
We Bulk-Reviewed 10 Vendor Contracts in 15 Minutes
We uploaded 10 vendor contracts simultaneously, ran parallel extraction, and got a comparison table that instantly showed which vendor had the worst terms β saving weeks of manual review.
Studio Founder
Comparing Two Contract Versions in 30 Seconds β Every Change Highlighted
When an investor sent back a revised term sheet, our AI agent diffed it semantically in 30 seconds β flagging who benefits from each change, not just what changed.
Studio Founder
Preparing Share Transfer Documents β Stock Transfer in 20 Minutes
We generate ordre de mouvement, certificat de cession, and registre des mouvements in French SAS format β handling prΓ©emption and agrΓ©ment procedures β in 20 minutes instead of 2 weeks.
Want results like these?
Start free with your own AI team. No credit card required.