US-Based Coupa Acquires UK's Rossum to Strengthen AI Document Processing, T-LLM Replaces Traditional OCR to Cover End-to-End Spend Management
2026-05-16 15:56
Favorite

en.Wedoany.com Reported - Autonomous spend management platform Coupa officially announced at its Inspire 2026 conference in Las Vegas that it has acquired Rossum, a market leader in AI intelligent document processing. The two companies have had a partnership since 2024, during which Rossum's document processing tools were embedded into the Coupa platform for accounts payable team invoice automation. This acquisition will extend Rossum's technology from accounts payable across Coupa's entire source-to-pay platform. Kirkland & Ellis served as legal advisor to Coupa, Guggenheim Securities served as exclusive financial advisor to Rossum, and Orrick served as legal advisor to Rossum.

The core element of the acquisition is Rossum's proprietary Transactional Large Language Model (T-LLM). This domain-specific model, built from the ground up for transaction document processing, has been trained on hundreds of millions of annotated transaction documents, covering invoices, purchase orders, logistics documents, and other categories. It represents a generational difference from general-purpose large models that only excel at continuous text understanding. T-LLM can automatically identify document layouts without templates, extract key fields, and provide confidence scores, while enforcing strict boundary controls on output values. Through a discriminative decoder, it ensures generated answers do not exceed the scope of the input document, thereby eliminating the value hallucination risk common in general-purpose LLMs at the architectural level. Compared to common optical character recognition systems in the industry, T-LLM reconstructs the document processing workflow from "extraction followed by manual correction" to "AI automatic extraction plus human review of exceptions," significantly reducing the manual intervention workload for finance teams.

Coupa CEO Leagh Turner described the transaction in the acquisition announcement as a strategic upgrade to a validated partnership. She noted that over the past 20 years, Coupa has saved customers over $300 billion, and with Rossum's technology, the next $300 billion could be achieved within 5 years. Rossum CEO and co-founder Tomáš Gogár viewed the acquisition as a natural evolution of the multi-year collaboration, stating that combining T-LLM's transactional intelligence with the $10 trillion community data covered by the Coupa network has the potential to fundamentally change how global enterprises buy and sell. Rossum was founded in 2017 by Tomáš Gogár, Petr Baudiš, and Tomáš Tunys, all three co-founders being artificial intelligence PhD dropouts. The company is headquartered in Prague, Czech Republic, and London, UK, has raised $100 million in venture capital, and serves over 275 blue-chip clients including Bosch, Panasonic, AB InBev, and Siemens.

The integration of T-LLM with Coupa's existing Navi agent fleet is the core technical rationale behind the acquisition. Rossum's transaction document processing capabilities cover multiple categories such as invoices, purchase orders, and logistics documents. Following the merger, Coupa plans to extend document processing functionality from accounts payable to the entire chain of direct and indirect spend. Previously, supplier document formats in direct spend scenarios varied greatly with uneven data quality, requiring traditional OCR solutions to expend significant manpower on template configuration and manual correction. T-LLM's template-free learning mechanism, combined with the autonomous decision-making capabilities of the Navi agent fleet, aims to achieve seamless connection from document reading to business decision-making across the entire source-to-pay process. The Coupa network connects over 10 million buyers and suppliers, and the community data accumulated on the platform provides a training foundation for T-LLM's continuous learning, allowing each customer's unique document sets to be absorbed by the model, accelerating time to value.

Business validation data supports the rationale for the acquisition. During the 2024 partnership, Rossum's embedded capabilities were validated by customers in the accounts payable scenario on the Coupa platform. Since the release of the Rossum Aurora engine in February 2024, early user data shows: a 10x reduction in the number of training documents required to achieve target accuracy, an average error rate reduction of 37.6% compared to the previous generation AI, and zero hallucination output. In public benchmarks such as DocILE, Rossum Aurora's F1 scores significantly led control models like GPT-4 and DocLLM-7B, validating its technical approach of "domain-specific outperforms general-purpose." Rossum also supports recognition of 276 languages and handwriting in 30 languages, with service coverage spanning multiple regional markets globally.

Founded in 2006 and headquartered in San Mateo, California, Coupa has developed, with backing from Thoma Bravo, into an autonomous spend management platform covering procurement, accounts payable, expense management, and supply chain planning. The current finance and procurement software market is witnessing accelerating consolidation around AI tools, with intelligent document processing becoming a key category for software vendors to reduce labor costs in document-intensive processes such as invoice management and supplier onboarding. Coupa's choice to acquire rather than continue an external partnership to internalize T-LLM as its own capability reflects a shift in the autonomous spend management sector's positioning of AI document processing from an "optional plugin" to "core infrastructure."

This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com