Contract Drafting Assistant

Procurement teams previously drafted formal contracts and appointment letters by manually parsing through dozens of PDF pages—official memos, vendor bids, and meeting minutes.

Staff had to hunt down scattered clauses, manually recalculate financial totals, verify vendor bank accounts, retype letter numbers, and translate numbers into Indonesian currency words (kalimat terbilang). The entire manual pipeline was slow, tedious, and prone to costly transcription errors on critical legal documents.

We architected a streamlined, human-in-the-loop pipeline:

1. Source Document Ingestion: Staff upload source PDFs directly into the platform.
2. Vision & LLM Extraction: An orchestration pipeline powered by vision OCR and structured LLM extractors parses all critical entities (vendor credentials, letter identifiers, breakdown of pricing, and bank accounts).
3. Side-by-Side Verification: Users review extracted parameters side-by-side with the original PDF on a dual-pane interface before committing any changes.
4. Automated DOCX Assembly: Standardized DOCX drafts are generated dynamically from centralized enterprise templates.
5. In-Browser Document Editing & Audit: Generated drafts can be polished in an integrated online document editor, with every validation and modification tracked in an audit log.

Procurement staff previously spent ~60 minutes manually cross-checking multi-page PDFs to retype vendor details, bank accounts, and currency words, resulting in typographical errors on 12–20% of initial drafts. Compressing turnaround to ~2 minutes through side-by-side verification and centralized templates is projected to save ~145 hours of labor per month (assuming 150 documents monthly, equivalent to 1 FTE, projected) while ensuring every modification is captured in an operational audit trail.