日本語 
Sansan Corporation, which provides AX services that change the way people work, has launched “Matching Assist,” a new feature for its accounting AX service “Bill One,” which uses AI to assist in matching invoices with delivery and acceptance data.
This function uses AI to identify variations in terminology, including synonyms and related terms, when comparing items listed on invoices with delivery and acceptance data from corresponding core systems, and presents items that it deems identical. This allows staff to perform the matching process using the results as a reference, significantly reducing the burden of manual verification.
Furthermore, we will begin building a process in which the results of verification and correction performed by staff members are accumulated as training data, allowing the AI to continuously learn. As a result, even in verification tasks where staff members previously judged matches or mismatches based on their experience, the AI will be able to accurately identify and present the results, eliminating the reliance on individual judgment.
Also Read: Ajike Launches FINUI Financial Design Service
This feature enhances the “AI Automatic Matching” that was launched in November 2025, and a patent application has been filed for it.
Many companies perform a monthly reconciliation process to verify that the contents of received invoices match the delivery and inspection data exchanged with suppliers at the item level. Because product information databases (product masters) differ from company to company, variations in terminology for the same product can occur between invoices and delivery/inspection data, such as “apple” and “apple” (in Japanese). Our verification of approximately 30,000 invoice lines revealed that more than half of the item lines contained such variations, and this tendency is particularly strong in industries such as wholesale and food, which have multi-layered supply chains. In some transactions, reconciling invoice data that can amount to thousands of lines requires personnel to visually check for variations in terminology and even minor differences in model numbers and specifications, which has been a significant burden.
Bill One developed this feature to eliminate the burden of matching data due to inconsistencies in notation and minor differences in model numbers and specifications.
SOURCE: PRTimes
English