ProcurePulse

Drop & Extract

Drop any vendor contract — PDF or Word. AI reads it in seconds, extracts every key field, and pre-fills your entire contract form. The fastest way to bring a vendor-paper contract into your system.

📄 Vendor PDF or Word
🔍 AI Page-by-Page Read
Fields Extracted
🔗 Vendor Matched
Form Pre-filled

Powered by Gemini 2.5 Flash vision + Claude Sonnet. Scanned PDFs supported via OCR (English, Hindi, Marathi).

What AI extracts

Every field. In seconds.

No templates to configure. No field mapping. Drop any contract — vendor-format, scanned, or typed — and the AI figures out the structure.

Contract title
e.g. Annual Maintenance Agreement for CNC Machines
Parties & roles
Buyer, Seller, or Service Provider — both sides identified
Contract type
Supply / Service / Rate Contract / Framework — auto-classified
Effective & expiry dates
Start date, end date, notice period
Contract value & currency
Total value, with currency (INR, USD, EUR)
Payment terms
Net-30, advance %, milestones, LC terms
Governing law & jurisdiction
State, country, court of jurisdiction
Dispute resolution
Arbitration clause, mediation, seat
Vendor master match
Fuzzy name match against your approved vendor list

How Drop & Extract works

Under the hood — transparently.

1

Drop the file

Drag a PDF or DOCX onto the new contract form. Any format accepted — typed documents, scanned copies (we OCR them), or Word files already in track-changes mode.

Supports English, Hindi, and Marathi OCR via Tesseract 4.
2

AI reads page by page

For PDFs, each page is converted to an image and processed in parallel batches (up to 8 pages concurrently) by Gemini 2.5 Flash vision. You see a live progress indicator — page 3 of 17 extracted. For Word files, Apache POI extracts the full text first.

Progress events stream live via Server-Sent Events — you see extraction happen in real time.
3

Fields structured by Claude

The raw extracted text goes to Claude Sonnet, which outputs a strict JSON object with every field. The JSON schema is enforced — no hallucination, no made-up values. Missing fields return null, not guesses.

OpenRouter → Claude Sonnet with json_schema + strict:true mode. Every field has a confidence band.
4

Vendor name fuzzy-matched

The vendor name from the contract (e.g. "Tata Consultancy Services Limited") is fuzzy-matched against your vendor master using Levenshtein distance. Likely matches are shown for one-click selection — no manual vendor lookup.

If no match found, you can create a new vendor record directly from the form.
5

Form pre-filled, you review

Every extracted field populates the contract creation form. You review, correct if needed, and submit. The AI does the 90% heavy lift — you do the 10% judgement call. Nothing is saved until you explicitly submit.

Section Map runs in parallel — by the time you review the form, clause analysis is already ready.

What happens next — automatically

Section Map scores every clause

After extraction, the Section Map parser runs on the uploaded document in parallel. By the time you finish reviewing the pre-filled form, the AI has already detected and scored every clause in the contract.

Indemnity 72% — deviates from standard
Limitation of Liability 91% — matches standard
Governing Law 100% — matches standard
Termination for Convenience 44% — review required
Force Majeure 88% — matches standard

Then — AI risk scan

Claude reads every clause for risk

After the Section Map, click "Run AI Audit" and Claude Opus analyses every clause for legal risk — uncapped liability, one-sided termination, missing IP ownership clauses, unfair indemnity. Each risk gets a severity, a quoted excerpt, and an explanation.

AI Risk finding — HIGH
"Clause 12.3 — Limitation of Liability"
"...liability shall not exceed ₹50,000 under any circumstances..."
Risk: Liability cap is disproportionately low relative to contract value (₹2.4Cr). In a breach scenario, recovery is limited to 2% of contract value. Recommend negotiating to minimum 100% of contract value or 12-month fees.
Accept Flag for Negotiation

Then negotiate — in Word, with real track changes

After extraction and risk scanning, the redline workflow begins. Both parties propose revisions. Every change generates a real OOXML Word file — not coloured highlighting, but genuine w:ins / w:del tracked-change XML that opens natively in Microsoft Word's Review pane.

Step 1

Download contract as Word

Platform generates a clean DOCX with Track Changes pre-armed. Vendor opens in Word, edits normally — their changes are tracked.

Step 2

Upload vendor's counter-draft

Vendor uploads their edited Word file. Platform extracts proposed changes, scores similarity, and shows the diff in the browser and as a downloadable redline DOCX.

Step 3

Accept or reject field-by-field

Buyer accepts or rejects each proposed change. Accepted changes apply to the live contract. SHA-256 canonical hash re-computed — any prior signatures become stale.

How this compares

Capability Most CLM tools Procol ProcurePulse
Import vendor contract PDF Manual retyping ✗ No CLM ✓ AI Drop & Extract
Scanned / image PDF support ✓ OCR (En, Hi, Mr)
Word track changes (real OOXML) ❌ Colour only ✓ w:ins / w:del
Vendor redline via portal ✓ Vendor portal
Clause risk scoring (AI) Paid add-on ✓ Claude Opus
Section Map — 22 clause types ✓ Built-in
Aadhaar eSign (Digio) ✓ IT Act §10A
SHA-256 signature integrity ✓ Hash-based
Obligation proposals from AI ✓ One-click apply

Trusted by GE, Tesco, Investec, Allergan, a global logistics leader, a Fortune 500 data storage manufacturer, Axis Bank, RBL Bank, and 500+ enterprises across 7 countries. Read customer stories →

See Drop & Extract live

30-minute demo — we'll drop a real vendor contract PDF and you'll watch the AI fill the form in real time.