Daily Japan updates
Three AI launches move from conversation into everyday workflows
New services from Resona, Light Up, and Knowledge Sense show Japanese businesses embedding AI into branch procedures, customer operations, and repeatable browser work.
Developments
Three announcements on September 14 point in the same direction: Japanese companies are moving AI beyond stand-alone chat and placing it inside repeatable operating processes. The changes cover customer-facing financial procedures, LINE-based sales operations, and browser work that can be handed from one employee to another.
1. Resona puts AI inside an operator-assisted branch counter
Resona Bank and Saitama Resona Bank launched a television counter service that combines a remote operator with AI support. During the conversation, the system organizes and captures information needed for a procedure; the customer then checks the details on screen instead of filling out a paper form. The service began at Shibuya, Abenobashi, and Kawagoe on September 14, with six more listed branches scheduled to follow by September 18. Resona says it aims to cut an average procedure from about 30 minutes to about 15, but that remains a company target rather than a measured outcome. (Read Resona’s announcement)
2. L-Sync opens customer-operation settings to MCP clients
Light Up added Model Context Protocol support to L-Sync, its AI customer-response service for LINE official accounts. The announced integration lets authorized AI clients or APIs update delivery steps, tags, rich menus, and scenarios, as well as read account information and delivery results. Importantly, the company says the adopting business still decides whether a proposed change is executed. That human approval point matters because the integration moves AI from drafting responses toward altering live operational settings. (Read Light Up’s announcement)
3. Cowork expands saved instructions from 2,000 to 100,000 characters
Knowledge Sense increased the maximum length of instructions stored in Cowork templates from 2,000 to 100,000 characters. Cowork is designed to save browser-based procedures so they can be repeated, and the larger limit allows teams to preserve branches, exceptions, and detailed steps instead of compressing a process into a short prompt. The announcement verifies the configuration change, not the accuracy or safety of the resulting automation, so businesses still need testing, access controls, and clear ownership of exceptions. (Read Knowledge Sense’s announcement)
What businesses should take away
The practical dividing line is no longer whether a company has an AI chat interface. It is whether the system can work inside a real process while preserving review, permissions, and a reliable fallback. For teams evaluating these products, the most useful questions concern who can authorize changes, which data the tool can reach, how exceptions are handled, and what evidence exists beyond the vendor’s launch announcement.