AI System for Automated Email-to-ERP Order Processing

Challenge

Our client is a procurement service that supplies various government agencies — from fire departments and industrial safety services to law enforcement and municipal utilities. The company delivers a wide range of products, from air cylinders and personal protective equipment to rescue tools and specialized equipment.

For emergency services, reliable supply chains are critical — their ability to respond quickly depends on how fast and accurately they can be equipped.

Historically, most supply requests were handled manually. Agencies would send requests via email, often in inconsistent formats: some data was in the email body, others in attachments (PDF or Excel).

The company processed dozens of such emails daily. Their staff had to manually extract order numbers, verify items, check delivery conditions, and transfer the data into their ERP system (NetSuite). This process was time-consuming, prone to human error, and did not scale well as the number of orders grew.

The client needed a system that could standardize request processing, automatically interpret their contents, route them through the appropriate workflows, and provide centralized control.

Solution

We developed a system that fully automates the processing of incoming order confirmations. It connects to corporate email via Microsoft Graph API, identifies relevant emails, and extracts data from attachments using OCR (powered by the Gemini 2.5 Pro model) or from plain text in the email body.

Our AI model extracts key data like order numbers, item lists, delivery dates, and addresses. These data points are then validated via the NetSuite API and automatically entered into the ERP system.

In case of discrepancies, the system logs them and generates a report for human operators, allowing them to focus only on exceptional cases. All core modules are deployed on Azure, enabling the system to scale with increasing data volumes without performance degradation.

Key features

• Seamless supplier experience: The system focuses on processing emails and attachments automatically without changing the supplier-side workflow — for suppliers, nothing changes,
• AI-based classification module: Routes requests into appropriate categories based on description,
• Integration with product catalogs and registries: Cross-checks items, filters by compliance and compatibility,
• Status management: Tracks the full request lifecycle — from approval to delivery confirmation,
• Integration with ERP and logistics systems: Auto-generates purchase or shipping orders;
• History and analytics: Provides full visibility into demand across regions, departments, and seasons.

The MVP was developed in just two months. The first three weeks were spent integrating with email services and implementing basic data extraction. The remaining time was used to connect to NetSuite and build the validation system.

Results

As a result, our client eliminated manual processing of emails and attachments. Order processing time was reduced from tens of minutes to just a few, and human error was virtually eliminated.

Procurement staff now focus on quality control and analytics, rather than copying data from emails into the ERP.

The solution is scalable and can be extended to support additional attachment formats and integrations with new suppliers, opening the door to further digitization of supply chain operations.

Digitizing procurement processes improved operational readiness, simplified document workflows, and increased transparency — a crucial benefit for agency oversight and future budget planning.