Darkmatter Logo

Store Intelligence: AI-powered contract and store operations hub

What started as a centralized lease contract tool evolved into an intelligent workspace with OCR extraction, contract alerts, store data management, and guided workflows.

0:00 / 2:01
Store Int 01

Challenge

A multi-country retailer needed a reliable way to manage lease agreements and store information at scale. Key contract details lived inside PDFs, spread across teams and folders, and required manual input and review. That created recurring problems:

  • Too many hours spent on contract input and analysis
  • Inconsistent contract data due to manual entry
  • Limited visibility into expiring contracts and key obligations
  • Slow search and retrieval across stores, contracts, and documents
  • Data quality issues across store records, with no clear workflow to fix them

Outcomes

We delivered an evolving product with two clear phases.

Phase 1: Store Manager Center (centralized contract management)

  • Intelligent OCR to extract contract data and pre-fill structured fields
  • LLM-based analysis flow to identify and extract requested clauses and fields
  • Validation steps to improve accuracy and reduce rework
  • Optimized rent contract workflow, plus fast filtering and search across contracts and stores

Phase 2: Store Intelligence (operational visibility and workflows)

  • Home dashboard with contract visibility and upcoming expirations
  • Store directory with structured attributes, status, and related contract tracking
  • Task system to operationalize data quality fixes and store updates (assignees, priority, status)
  • Documents area to centralize supporting files and improve auditability
  • An assistant-style entry point for search and guided actions across the workspace

Tech Stack

Data Extraction: Mistral OCR, Google Gemini, Python

Data Management: Python, PostgreSQL

Frontend: React, Next.js

FAQs

Phase 1 delivered working contract extraction within weeks. The system started reducing manual input time immediately, with accuracy improving as the validation feedback loop trained the extraction models.

The system uses Mistral OCR combined with LLM-based analysis to handle varying contract layouts. It learns from corrections and adapts to new formats without requiring manual template configuration.

Yes. The platform is designed to work alongside existing tools. Data can be exported or synced to ERP systems, spreadsheets, or other contract management tools as needed.

Users review and correct extractions through the web interface. These corrections feed back into the system, improving future accuracy. The validation workflow ensures nothing goes unreviewed.

The system was built for lease contracts but the underlying approach works for any structured document extraction. The same architecture can handle supplier contracts, invoices, or other document types.

Related

How a global retailer eliminated manual invoice processing with automated data extraction

How a global retailer eliminated manual invoice processing with automated data extraction

Read more