Sansan Implements 'Cello' AI Model Developed in National 'GENIAC' Project into Its Products
Sansan, Inc. has integrated 'Cello,' a document-specialized AI model developed through the national GENIAC project, into its services 'Contract One' and 'Bill One.' Cello enhances data accuracy and processing speed by outputting text and positional data simultaneously.
📋 Article Processing Timeline
- 📰 Published: May 19, 2026 at 20:00
- 🔍 Collected: May 19, 2026 at 11:31
- 🤖 AI Analyzed: May 21, 2026 at 15:48 (52h 16m after Collected)
## Sansan Implements 'Cello' AI Model Developed in National 'GENIAC' Project into Its Products
Sansan, Inc., a provider of Analog Transformation (AX) services that change the way people work, has announced the implementation of 'Cello,' a document-specialized AI model developed under the 'GENIAC' (Generative AI Accelerator Challenge) project. The project is conducted by the Ministry of Economy, Trade and Industry (METI) and the New Energy and Industrial Technology Development Organization (NEDO) to strengthen Japan's generative AI development capabilities.
This model has been integrated into the data processing workflows of Sansan's contract management service 'Contract One' and accounting AX service 'Bill One.' This integration further improves the accuracy and speed of data conversion for business documents such as contracts and invoices, contributing to enhanced user operational efficiency and productivity. Plans are underway to implement Cello in other products, including the business database 'Sansan,' in the future.
### About the 'Cello' Document-Specialized AI Model
'Cello' is a proprietary AI model built by Sansan during the third phase of the national GENIAC project, which supports the development of foundational generative AI models.
Sansan has previously developed 'Viola,' a multimodal generative AI that extracts necessary items directly as text by using question-answering formats for document images. 'Cello' is an advanced version of 'Viola.' It features the capability to output positional information, indicating where the extracted text is located on the document, alongside the extracted text itself. This evolution allows the AI to process information while identifying the text regions that serve as the basis for its answers, thereby enabling the construction of a mechanism to detect and eliminate hallucinations—a common challenge for generative AI. This ensures high consistency with Sansan's data conversion rules, enabling extremely accurate data processing.
Furthermore, in Sansan's data conversion workflow, 'Cello' integrally processes steps that were previously divided, such as 'character data conversion' and 'item identification and meaning understanding.' This improves the processing speed of the entire data pipeline, allowing for faster data delivery.
### The Importance of Data in Business AI and Sansan's Approach
As the use of generative AI becomes widespread, relying solely on publicly accessible information makes it difficult to generate a competitive advantage. Incorporating and leveraging 'primary information' based on a company's unique facts is the source of competitive advantage. However, AI performance in business is highly dependent on data quality. Since AI uses input information as its judgment standard, noisy unstructured data becomes a cause of hallucinations. Therefore, it is essential to prepare structured data that eliminates guesswork and allows AI to process accurately.
Sansan has established a unique data processing system that combines AI technology using machine learning, strict data conversion rules, and human quality assurance. This operational foundation, which creates highly accurate data from diverse analog documents, has become a strong competitive advantage in the modern era where AI is prevalent. The implementation of 'Cello' marks the next stage in the evolution of this long-cultivated data operation.
By implementing this model, Sansan aims to further increase corporate productivity and strengthen its high-precision database foundation for maximizing AI utility.
Sansan, Inc., a provider of Analog Transformation (AX) services that change the way people work, has announced the implementation of 'Cello,' a document-specialized AI model developed under the 'GENIAC' (Generative AI Accelerator Challenge) project. The project is conducted by the Ministry of Economy, Trade and Industry (METI) and the New Energy and Industrial Technology Development Organization (NEDO) to strengthen Japan's generative AI development capabilities.
This model has been integrated into the data processing workflows of Sansan's contract management service 'Contract One' and accounting AX service 'Bill One.' This integration further improves the accuracy and speed of data conversion for business documents such as contracts and invoices, contributing to enhanced user operational efficiency and productivity. Plans are underway to implement Cello in other products, including the business database 'Sansan,' in the future.
### About the 'Cello' Document-Specialized AI Model
'Cello' is a proprietary AI model built by Sansan during the third phase of the national GENIAC project, which supports the development of foundational generative AI models.
Sansan has previously developed 'Viola,' a multimodal generative AI that extracts necessary items directly as text by using question-answering formats for document images. 'Cello' is an advanced version of 'Viola.' It features the capability to output positional information, indicating where the extracted text is located on the document, alongside the extracted text itself. This evolution allows the AI to process information while identifying the text regions that serve as the basis for its answers, thereby enabling the construction of a mechanism to detect and eliminate hallucinations—a common challenge for generative AI. This ensures high consistency with Sansan's data conversion rules, enabling extremely accurate data processing.
Furthermore, in Sansan's data conversion workflow, 'Cello' integrally processes steps that were previously divided, such as 'character data conversion' and 'item identification and meaning understanding.' This improves the processing speed of the entire data pipeline, allowing for faster data delivery.
### The Importance of Data in Business AI and Sansan's Approach
As the use of generative AI becomes widespread, relying solely on publicly accessible information makes it difficult to generate a competitive advantage. Incorporating and leveraging 'primary information' based on a company's unique facts is the source of competitive advantage. However, AI performance in business is highly dependent on data quality. Since AI uses input information as its judgment standard, noisy unstructured data becomes a cause of hallucinations. Therefore, it is essential to prepare structured data that eliminates guesswork and allows AI to process accurately.
Sansan has established a unique data processing system that combines AI technology using machine learning, strict data conversion rules, and human quality assurance. This operational foundation, which creates highly accurate data from diverse analog documents, has become a strong competitive advantage in the modern era where AI is prevalent. The implementation of 'Cello' marks the next stage in the evolution of this long-cultivated data operation.
By implementing this model, Sansan aims to further increase corporate productivity and strengthen its high-precision database foundation for maximizing AI utility.
FAQ
What is Sansan's 'Cello'?
A high-precision, document-specialized AI model developed via the GENIAC project.
Which products include Cello?
It is implemented in the data processing flow of Bill One and Contract One.
What are the benefits of Cello?
It enables accurate and fast data extraction by processing both text and positional information.