How Wolt established scalable invoice automation across 11 countries with 60% straight-through processing

Who is Wolt?

Wolt is a leading technology company that provides a platform for food and retail delivery services across Europe. With operations in 31 countries and rapid expansion of the Wolt Market online grocery stores, it faces complex document processing challenges across multiple languages, legislations, and vendor types. Its AP team processes 45-50,000 invoices monthly from vendors ranging from major suppliers to small local businesses.

Business objective

Wolt wanted to establish a centralized AP function to process invoices from all Wolt Market operating countries, with volumes growing 30% a year. To support efficiency goals, the team started looking for a scalable solution for its rapidly growing invoice volume without exponentially increasing headcount.

Specific goals included:

• Providing analytical data for operational teams to optimize inventory management.
• Establishing three-way matching to reconcile purchase orders with invoices and ensure fair payment to suppliers.
• Processing high volumes of invoices efficiently across multiple countries and languages.

Solution requirements also included ease of integration with Wolt’s downstream order management platform.

“To support our growing business, we needed one system to ensure compliance, build trust, improve our AP process, and reconcile information effectively.”

Jeļena Bičková, Wolt Market AP Process Owner

What was the challenge?

Wolt faced several significant challenges in its document processing:

• The existing OCR solution could not handle line item data extraction.
• Handling invoices in Czech, Slovak, German, Estonian, Greek, Hungarian, Lithuanian, Latvian, and Slovenian without AP specialists speaking those languages.
• Many small vendors send handwritten or poorly formatted invoices (or missing critical information like VAT).
• Different tax regulations and document formats across 31 countries.
• High risk of overpaying suppliers because of no three-way matching on line item level.
• No analytical data available to understand customer behavior, stock variances.

These issues affected Wolt Market’s margins and prevented data-driven business development.

The solution: Scalable AI-powered invoice processing

Wolt implemented Rossum’s AI-powered intelligent document processing solution to address their complex invoice processing needs. The solution ingests invoices via email, extracts and validates data, and pushes it to their internally developed fulfillment tool before final export to NetSuite.

Key features include:

• AI-powered data extraction that handles multiple languages and document formats.
• Customized field extraction based on country-specific requirements.
• Validation rules including amount checksums, line item vs. header amount checks, and tax validations.
• Data transformation and inference for generating VAT rates, due dates, and calculating quantities.
• Live document translation to enable centralized processing of documents in various languages.

Results

After going live in 2024, Wolt has since scaled the Rossum implementation to 11 countries while keeping the same headcount. It takes 3-6 weeks to onboard a new country depending on the market size.

“We can constantly push automation higher, so that we can reassign resources to new countries where we want to launch Rossum.”

Bence Simon, AP team lead, Wolt

Wolt has achieved 60% automation on average so far, with a median document review time of 47 seconds.

Besides that Wolt has achieved the goals that were crucial from the beginning:

• Establishing three-way matching and ensuring they are not overpaying their suppliers.
• Providing reliable data to operational teams about stock variances and enabling them to manage inventory effectively.

The future: scaling Rossum across geographies

Wolt continues to expand their use of Rossum as part of its global AP strategy. Immediate plans include:

• Onboarding additional countries, including Cyprus, Malta, Sweden, and Denmark
• Continuing to optimize its three-way matching process