Domestic AI Agent Trends (May 19, 2026 Issue)

Update Date: 2026/5/19

Executive Summary
The domestic AI agent trends for May 18, 2026, show implementation advancing across diverse business domains, including advertising operations, banking inquiries, management decision-making, financial research, meeting information utilization, accounting, PR effectiveness measurement, manufacturing maintenance, restaurant reservations, and game generation. What they have in common is that AI is beginning to handle not just answering, but also external data acquisition, follow-up questioning, estimation, evidence presentation, system integration, and execution support. The domestic market is shifting from a demonstration phase to a full-scale adoption phase where it is being integrated into business processes.

*I input the created article content into Gamma to have it automatically create slides. If you find slides easier to view, please take a look here.

1️⃣ Opt automates operational workflows with the advertising operation platform "OPT Ad Agent"

📎 Source: Digital Holdings, Inc. Press Release
Opt, Inc. has introduced the Claude Enterprise Plan company-wide and has begun full-scale structural transformation of its advertising operations based on AI agents. The proprietary "OPT Ad Agent" integrates with advertising media such as Google, Meta, LINE Yahoo, and TikTok, as well as Gmail, Slack, Figma, and Canva, to autonomously perform tasks ranging from data acquisition, aggregation, anomaly detection, and report creation to ad submission file generation and banner composition proposal creation. Aiming to shorten the PDCA cycle that previously took several weeks, the design shifts to one where humans focus on strategic dialogue and customer support.

2️⃣ Mizuho Bank to deploy conversational AI agent "Aomaru Bank" to banking app

📎 Source: Mizuho Bank, Ltd. News Release
Mizuho Bank announced that it will begin offering the conversational AI assistant "Aomaru Bank" in September 2026. The first phase is an inquiry response function on the Mizuho Direct app, which grasps and judges the user's situation and needs during the conversation, and asks follow-up questions to guide them to the optimal information if questions are vague or prerequisites are missing. By combining OpenAI's technology, voice AI, and FAQ/manual search, it ensures the accuracy and safety essential for financial services through evidence-based responses and multi-layered detection and blocking.

3️⃣ Mitsubishi Research Institute provides end-to-end support for business impact assessment with AI agents

📎 Source: Mitsubishi Research Institute, Inc. Press Release
Mitsubishi Research Institute has added a quantitative business impact assessment support function using AI agents to its "Intelligence Platform" for companies. Regarding the impact of external environmental changes on sales, profits, production volume, procurement, and supply chains, it executes everything from scenario creation, coefficient setting, and rough estimation to evidence presentation in an end-to-end manner. The output comes in three patterns: pessimistic, neutral, and optimistic, emphasizing the explainability and persuasiveness necessary for decision-making. Internal verification has confirmed a 9.2-point improvement in accuracy compared to chat-based generative AI alone.

4️⃣ AlpacaTech enables financial AI agent PoC with "MixSeek AI Data Lake" for Japanese stocks

📎 Source: FOLIO Holdings, Inc. Press Release
AlpacaTech built the "MixSeek AI Data Lake," a PoC platform for Japanese stock AI agents, on May 15, referencing the "financial-services" reference implementation for AI agents in the financial industry published by Anthropic. It has established an AI-Ready data infrastructure that allows AI agents to refer to public information such as TDnet's financial results summaries and timely disclosures, and EDINET's securities reports at high speed and scale. PoCs for seven skills have become possible: financial analysis, morning note creation, cross-sector analysis, investment hypothesis verification, competitor analysis, pre-analysis of the next financial results, and investment idea extraction.

5️⃣ Sourcenext begins offering "AutoMemo Copilot Agent" that runs on Microsoft 365 Copilot

📎 Source: Sourcenext Corporation News Release
Sourcenext began offering the "AutoMemo Copilot Agent," which can be used on Microsoft 365 Copilot, as a new feature of the AI meeting minutes service "AutoMemo" on May 18. It links with an AutoMemo account, allowing users to search, summarize, and extract information from accumulated past meeting data using natural language. A key feature is that users can find relevant meetings from topics or remarks even without remembering exact dates or file names, and can complete tasks such as drafting meeting minutes and extracting decisions, next actions, and specific remarks within Copilot.

6️⃣ TOKIUM introduces "TOKIUM AI Detail Input" to FORESTON to automate journal entry tasks

📎 Source: TOKIUM, Inc. Press Release
TOKIUM announced that FORESTON, which operates a nature tour business in Oirase Stream, Aomori Prefecture, has introduced "TOKIUM AI Detail Input." At FORESTON, invoices in various formats such as paper and PDF arrived from over 50 suppliers, and journal entry work was concentrated on one person in charge. By having AI learn past corrections and journal entry rules for each business partner, it suppresses reliance on specific individuals and expects to reduce work by about 20 hours per month. This is a case where accounting AI agents are penetrating the back-office automation of small and medium-sized businesses.

7️⃣ Todo-Onada to offer PR effectiveness measurement AI agent "Qoddler" on Qlipper

📎 Source: Todo-Onada, Inc. Press Release
Todo-Onada, Inc. announced that it will begin offering the AI agent function "Qoddler" on the PR effectiveness measurement service "Qlipper" starting May 25, 2026. In response to natural language instructions such as "Summarize yesterday's media exposure," "Compare the number of placements with competitors," and "Create a search formula to investigate new products," the AI autonomously investigates and displays information within Qlipper. By having the AI take over moving between multiple screens and setting conditions, it realizes an environment where PR personnel can focus on essential tasks such as strategy planning and competitor research.

8️⃣ Narecom adds API integration to "AI Agent Studio" to strengthen business system connectivity

📎 Source: Knowledge Communication Inc. Press Release
Knowledge Communication has added an "API Integration" feature to its AI agent construction tool for non-engineers, "KNOWLECOM AI Agent Studio," allowing it to link with external Web APIs. This enables the incorporation of HTTP-based API calls during agent processing, making it easier to handle tasks such as data retrieval, registration, updates, approval flows, and ticket creation for core systems and SaaS. The no-code UI allows frontline departments to design business workflows through trial and error, with integration into existing authentication and network policies as a prerequisite.

9️⃣ Accenture, Avanade, and Microsoft Co-develop "Agentic Factory"

📎 Source: Accenture Japan Ltd. News Release
Accenture and Avanade, in collaboration with Microsoft, have co-developed an "Agentic Factory" to help manufacturing companies reduce downtime. When production lines or machinery fall below expected operating rates, AI agents assist with initial checks, diagnostics, troubleshooting, root cause analysis, and the presentation of recommended countermeasures. If necessary, they can also issue maintenance tickets and prepare spare parts orders. Utilizing Microsoft Azure, Fabric, Foundry, and Copilot, the effectiveness of this solution is currently being verified by early adopters such as Kruger.

🔟 Tokyu Land Corporation x SoftBank Conduct Field-Led "AI Agent Contest"

📎 Source: SoftBank Corp. News Release
SoftBank, in partnership with Tokyu Land Holdings and Tokyu Land Corporation, co-hosted their first internal "AI Agent Contest." Targeting all employees, the contest solicited plans and development for "AI agents that solve problems experienced by the employees themselves," resulting in over 100 ideas. Proposals with high practical utility, such as legal contract checks, customer needs analysis for sales activities, and the advancement of development tasks, were prominent. This is attracting attention as a case study of bottom-up DX where frontline employees define how AI is used.

1️⃣1️⃣ NEIGHBOR Releases "DreamCore V2" Equipped with AI Agents Specialized for Game Production

📎 Source: NEIGHBOR Inc. Press Release
NEIGHBOR has released "DreamCore V2," a new version of its AI game platform "DreamCore" that implements proprietary AI agents specialized for game production. By simply giving instructions in natural language about the game they want to create, users can integrally generate code, character images, 3D models, sounds, UI, and even smartphone/PC operation support, allowing them to play and share the game immediately in a browser. While game generation may appear consumer-oriented, it is also being used for corporate training, program promotion, and AI experience classes, expanding the use of agents in creative and educational fields.

1️⃣2️⃣ Upsell Technologies Launches AI Operator "miraio" for Restaurants at Two Shinobuya Locations

📎 Source: Upsell Technologies Inc. Press Release
Upsell Technologies began providing "miraio," an AI operator co-developed and operated with pluszero, at Horii Food Service's "Kakure-an Shinobuya" Warabi Station East Exit and Yashio Station Front branches on May 18, 2026. It handles reservation reception, changes/cancellations, vacancy checks, reservation information verification, hearing names and phone numbers, course guidance, and FAQs, reducing opportunity loss through 24/7 automated responses. Using pluszero's AEI technology to suppress hallucinations, the goal is to create an environment where store staff can focus on face-to-face customer service.

Comprehensive Analysis

The features observed from the topics on May 18, 2026, indicated that AI adoption by domestic companies is clearly progressing from "trialing generative AI" to "implementing task-execution agents." As seen in the cases of Opt, Mizuho Bank, and Mitsubishi Research Institute, AI is becoming more than just a tool for information retrieval or text creation; it is becoming an entity that references business data, asks users follow-up questions, organizes decision-making materials, and connects with existing systems to move business forward. On the other hand, issues such as incorrect answers, permission management, and auditability are significant in finance, accounting, manufacturing, and customer support. Moving forward, comprehensive capabilities—including not just accuracy, but also safety design, accountability, and operational governance—will determine competitive advantage.

Future Points of Interest

• The focus of AI agent adoption is shifting from the convenience of the chat screen to how safely actual processing can be entrusted by connecting to existing data and business systems.

• Mizuho Bank's "Aomaru Bank" is attracting attention for both its response accuracy and risk control as a case study of a financial institution fully implementing conversational AI at customer touchpoints.

• Vertical AI agents that understand industry-specific procedures and data structures, such as in advertising, accounting, PR, and restaurant reservations, are likely to become a differentiator from general-purpose AI.

• As shown by the cases of KNOWLECOM and Opt, business connection design that combines APIs, MCP, Copilot, Claude, etc., will be a key theme that determines the effectiveness of adoption.

• The contest-style initiative by Tokyu Land Corporation and SoftBank has the potential to spread as a practical model for bottom-up DX where frontline employees plan AI agents themselves.