Domestic AI Agent Trends (November 29, 2025 Issue)
Update Date: 2025/11/29
💡 Executive Summary
As of November 28, 2025, the Japanese AI agent market is showing a decisive shift from the "proof-of-concept phase" to the "production implementation phase." Three major trends have become clear.
🎯 Key Trends
Full-scale deployment of industry-specific agents: Acceleration of integration into actual operations in fields such as healthcare (Ubie), finance (PKSHA x Funds), and customer support (Channel Talk, KDDI).
Polarization and convergence of administrative DX: The Digital Agency's in-house platform "Gennai" and proof-of-concept experiments by local governments are proceeding in parallel.
The barrier from PoC to production operation: An NRAC survey highlights the challenge of a production operation rate of approximately 10%. Meanwhile, major companies like KDDI and Nitto Denko have demonstrated quantitative results.
1️⃣ PKSHA x Funds: AI Agent for Financial Screening Operations
Source URL: https://prtimes.jp/main/html/rd/p/000000231.000022705.html
Summary: PKSHA Technology and Funds have entered into a capital and business alliance to jointly develop AI agents specialized for corporate finance operations (screening and monitoring). By automating information gathering and decision-support processes that tend to rely on individual expertise in financial practice, they are promoting social implementation that dramatically improves the quality and speed of operations through human-AI collaboration. Introducing agents into areas that require complex financial regulations and high levels of expertise is an important milestone in Japan's financial DX.
⚠️ Points of Note
The key is how far AI agents can handle the "reliance on individual expertise" and "complexity of judgment" unique to the financial industry. Clarifying governance and responsibility will be a future challenge.
2️⃣ Channel Talk "ALF v2": Toward Full Automation of Customer Support
Source URL: https://prtimes.jp/main/html/rd/p/000000142.000029184.html
Summary: Channel Corporation's chat service "Channel Talk" has significantly updated its AI agent "ALF." Evolving from conventional RAG-based inquiry responses, it has begun production deployment of "ALF v2," which automatically executes actual operational tasks such as order cancellations and return processing. With a goal of an 80% inquiry resolution rate, the role of the agent is expanding from simple information provision to "operational delegation." It is attracting attention as a practical solution to labor shortages in the customer support industry.
💡 Insight
A case symbolizing the evolution from "conversational AI" to "task-executing agents." We have entered an era where linkage with business systems (API integration) determines the practicality of agents.
3️⃣ Inazawa City, Aichi Prefecture x Graffer: Proof-of-Concept Experiment for AI-based Local Government Counter Operations
Source URL: https://prtimes.jp/main/html/rd/p/000000114.000038525.html
Summary: Inazawa City, Aichi Prefecture, has begun a proof-of-concept experiment for telephone response in waste disposal guidance operations using Graffer's "AI Operator." Due to the AI agent's advanced conversational capabilities, it is expected to reduce telephone response work by staff by 50%. Amidst difficulties in securing staff due to population decline, it is expected to serve as an AI utilization model for local governments that balances improved resident services with reduced operational burdens. It holds the potential for horizontal expansion as a solution that can be introduced even by small-scale local governments.
🏛️ Perspective on Administrative DX
The Digital Agency's "Gennai" (central government) and individual implementations by local governments are proceeding in parallel. It is highly likely that integration and standardization of both will progress in 2026.
4️⃣ E-Agency: Marketing Process Automation AI Agent (Beta Version)
Source URL: https://prtimes.jp/main/html/rd/p/000000439.000009812.html
Summary: E-Agency has incorporated two types of AI functions—a problem-discovery agent and a strategy-planning agent—as a beta version into its marketing campaign management platform, "AD EBiS Campaign Manager." The scope has expanded from supporting conventional planning processes to extracting issues based on variance analysis between goals and results, and autonomously planning improvement measures. By AI-enabling marketing analysis and planning tasks that tend to rely on individual expertise, it supports marketers in focusing on strategic work. It aims to improve quality toward the official version after verification in the beta version.
5️⃣ KDDI: Production Introduction of AI Agents for Chat Support Operations
Source URL: https://enterprisezine.jp/news/detail/23236
Summary: KDDI and KDDI Research have officially deployed their self-developed AI agent into 'au Chat Support.' Equipped with hallucination (incorrect answer) suppression technology, they anticipate approximately 90% response accuracy and a 70% reduction in response time. By streamlining 160,000 monthly staff-handled tasks, this serves as an advanced case study for AI in large-scale customer support operations within telecommunications carriers. The in-house development approach using proprietary technology is attracting attention.
📈 Visualization of Quantitative Effects
Presenting specific KPIs such as 90% response accuracy and a 70% time reduction serves as a key leading indicator for AI agent investment decisions. It becomes the benchmark for ROI calculation.
6️⃣ Nitto Denko x IBM: Full-scale operation of 'AI First BPO' in manufacturing begins
Source URL: https://jp.newsroom.ibm.com/2025-11-26-Nitto-with-ai-first-bpo
Summary: Nitto Denko has partnered with IBM to launch 'AI First BPO,' which automates 90% of expense reimbursement check tasks using agentic AI. As a large-scale automation case study for routine tasks in the manufacturing industry, it demonstrates a collaboration model between humans and AI. By fusing BPO (Business Process Outsourcing) with AI agents, they have established a new service model that achieves both efficiency and quality improvement in back-office operations.
7️⃣ Ubie: Special generative AI package for medical DX launched
Source URL: https://prtimes.jp/main/html/rd/p/000000190.000048083.html
Summary: Ubie, Inc. has begun offering a special package of 'Ubie Generative AI' and 'Ubie DPC Supporter,' including support for utilizing subsidies. By introducing AI agents to DPC (Diagnosis Procedure Combination) coding, a highly specialized task in medical institutions, Kameda Medical Center reduced information gathering time for cancer registration by approximately 30% annually, and Nanbu Tokushukai Hospital created 200 hours of monthly work time for creating informed consent records. Practical improvements through medical-specialized agents have been proven with data.
🏥 The Uniqueness of Medical AI
The key in the medical field is not full automation, but a 'Human-in-the-Loop' design. The 'augmented intelligence' model, where experts make final decisions while AI handles cognitively demanding processes, is becoming the standard.
8️⃣ Mukuil 'AI Success Partners': Field-oriented support to eliminate PoC fatigue
Source URL: https://www.logi-today.com/880338
Summary: Mukuil has launched 'AI Success Partners (AISP),' a field-oriented AI implementation support business. To address 'PoC fatigue,' the biggest challenge in AI adoption for Japanese companies, they provide comprehensive hands-on support including smart glass AI and on-site data analysis solutions. Their implementation approach emphasizes compatibility with on-site workflows for AI agent utilization in environments with physical entities, such as logistics warehouses and manufacturing sites.
9️⃣ BCG/MIT Survey: AI agent adoption rate at 35%, becoming close to colleagues
Source URL: https://www.bcg.com/ja-jp/press/27november2025-agentic-ai-blurs-line-tool-teammate
Summary: A global survey by Boston Consulting Group and MIT SMR (2,102 people across 116 countries and 21 industries) found that the AI agent adoption rate has reached 35%, with 44% planning to adopt. Corporate adoption is progressing faster than traditional AI (72% in 8 years) and generative AI (70% in 3 years). 76% of respondents perceive AI agents as 'close to colleagues,' confirming a shift in perception from tool to collaborative partner. However, risks of implementing before strategy and organizational preparation are also pointed out.
🔟 NRAC Survey: The reality of Japan's approximately 10% production operation rate
Source URL: https://note.com/nrac/n/n6c2ade82b022
Summary: The Next-generation RPA & AI Consortium (NRAC) raised the issue at 'AI Agentic Summit 2025' and in the 'Next-generation RPA & AI White Paper 2025' that the rate of Japanese companies moving from AI agent PoC to production operation remains at about 10%. Action execution, process design, and orchestration have emerged as major barriers. The necessity of business flow redesign and organizational transformation from an agent perspective is emphasized to address the structural challenges unique to Japan, where PoCs proceed but implementation lags.
⚡ Structural Challenges in the Japanese Market
The gap between the global adoption rate of 35% (BCG survey) and Japan's production operation rate of 10% (NRAC survey) suggests differences not in technical capability, but in 'business process redesign' and 'organizational culture'.
🎯 Integrated Insights: 3 Major Trends
Trend 1: The Victory of Vertical Agents
Agents deeply embedded in specific domain knowledge and business flows, such as Ubie (medical), PKSHA x Funds (finance), and Channel Talk (CS), are outperforming general-purpose AI and creating substantial value. General-purpose LLMs are the 'engine,' and the source of competitiveness is shifting to the 'chassis' (workflow design and proprietary data).
Trend 2: The Spread of Embedded Agents
As seen in Mukuil's smart glasses and Dify's trigger functions, AI is jumping out of chat screens and being embedded behind business systems and physical devices, evolving into 'Invisible AI' that users benefit from without even realizing it. We have entered an era where resident programs operating 24/7 orchestrate entire business processes.
Trend 3: Overcoming the 'Valley of Death' from PoC to Production is the Top Priority
The reality of an approximately 10% production operation rate revealed by the NRAC survey highlights challenges in 'business process redesign,' 'responsibility and governance structures,' and 'ROI visualization,' rather than technical maturity. Meanwhile, leading companies like KDDI and Nitto Denko are proving quantitative effects and establishing investment decision criteria.
📊 Outlook for 2026
Prediction 1: Acceleration of Administrative AI Integration
With the rollout of the Digital Agency's 'Gen-nai' to all ministries in January 2026, the integration of AI infrastructure between the central government and local municipalities will proceed. Standardization of administrative services and the establishment of an AI governance framework are expected.
Prediction 2: Establishment of Regulatory Compliance Models in Finance and Healthcare
Leading cases in highly regulated industries, such as PKSHA x Funds and Ubie, are forming best practices for the division of responsibility and governance structures in AI agent usage. The development of specialized insurance products is also coming into view.
Prediction 3: The Rise of Physical AI Agents
With the use of smart glasses by Mukil and the inheritance of tacit knowledge by NTT Data, the use of agents in physical spaces is becoming full-scale. As 'digital mentors' in manufacturing and logistics sites, they have the potential to become a solution to the crisis of passing on skilled expertise.
