Huliot automates multilingual purchase orders with DocuPipe

For manufacturers serving a diverse customer base, purchase order processing is a constant bottleneck. Every order that arrives by email must be converted into a sales order before production can be scheduled. The faster that conversion happens, the faster product ships.

But purchase orders don't arrive in a standard format. They come from dozens of different ERP systems: SAP, Oracle, Priority, and proprietary systems built in-house by customers. Each system exports purchase orders differently. Field names vary. Layouts differ.

Huliot, a plastic piping manufacturer headquartered in Israel with operations across Europe and India, faced this at scale. With multiple business units and customers across multiple countries, thousands of purchase orders arrived monthly via email.

The documents required extraction in Hebrew, a right-to-left language that most OCR systems handle poorly. Huliot's previous workflow relied on ABBYY for basic OCR combined with staff manually entering data into their Priority ERP system. The process was slow, error-prone, and couldn't scale with business growth.

Huliot evaluated multiple solutions but found that most document processing tools fail on non-Latin scripts. Hebrew's right-to-left text direction broke most tools they tried.

During proof-of-concept, DocuPipe extracted data from Huliot's most challenging formats: Hebrew purchase orders with handwritten annotations, poor scan quality from faxes. DocuPipe handled all of them.

DocuPipe's team worked directly with Huliot to build and tune the initial extraction schemas — handling complex business logic like dual SKU extraction (customer SKU and supplier SKU), branch identification from PO number prefixes, currency detection across shekels, dollars, and euros, and customer-specific rules for how fields should be interpreted. The onboarding was hands-on: document samples were reviewed together, edge cases were identified, and schemas were iterated until extraction met production standards.

From there, the implementation started with the highest-volume business unit and expanded across the organization. Email inboxes that previously required manual monitoring now feed directly into the automated pipeline.

The Hebrew language support that seemed like the hardest requirement turned out to be straightforward. DocuPipe handles Hebrew natively. The extraction works with common patterns in Israeli business documents: Hebrew addresses, phone numbers, and date formats.

DocuPipe's team built the initial schemas for Huliot's most complex document types — but the goal was always to hand over the keys. Once the foundation was in place, Huliot's IT team took over day-to-day management: refining schemas, adding new document types, and expanding to additional business units on their own.

Today, Huliot manages their extraction pipeline independently — iterating schemas and configuring workflows without needing to file support tickets or wait on DocuPipe for changes.

Once an order completes extraction, a webhook triggers a Make.com integration that maps the extracted data into Huliot's Priority ERP. Purchase order data flows from email to ERP without human intervention. What used to take hours or days during peak periods now takes minutes.

Huliot's IT team configured the data mapping themselves using DocuPipe's webhook output and Make.com's workflow builder. No custom development required. No integration consultants.