Domestic AI Agent Trends (December 20, 2025 Issue)
Update Date: 2025/12/20
📊 Executive Summary
The domestic AI agent market as of December 19, 2025, demonstrates that AI agents have evolved from "support tools" to "autonomous proxies." This is evidenced by concrete achievements such as the establishment of a public-private consortium by Osaka Prefecture, the launch of a multi-agent platform by Dentsu Digital, and a 40-point improvement in goal achievement rates through Neocareer's sales AI.
1️⃣ Establishment of the Osaka Prefecture Administrative AI Agent Consortium: A New Horizon for Public DX
Source URL:
🏛️ Summary
Osaka Prefecture has established the "Osaka Prefecture Administrative AI Agent Consortium." Approximately 20 industry-government-academia organizations, including AWS, Google Cloud, Microsoft, NTT Group, Salesforce, and Nvidia, are participating, with full-scale implementation targeted for fiscal year 2027. The initiative focuses on reducing administrative burdens for school staff, interactive proxy handling of administrative procedures, and multilingual support, elevating AI to the "front desk" of administrative services. With the participation of Sky Co., Ltd., implementation capabilities for educational settings are also secured. It has high potential to become a model case for local governments nationwide.
⚡ Strategic Impact
Local governments are shifting from standalone AI utilization to an ecosystem-based development structure in collaboration with global companies. By utilizing a multi-cloud environment, they avoid vendor lock-in while enabling the selection of optimal LLMs for specific purposes. If the "Osaka Model" succeeds, it will become the standard for administrative AI in Japan.
2️⃣ Neocareer: Dramatic Improvement in Goal Achievement Rate from 50% to 90% with Sales AI
Source URL: https://prtimes.jp/main/html/rd/p/000000231.000097462.html
📈 Summary
Neocareer Co., Ltd. introduced the "Sales Marker" meeting recording and AI multi-agent analysis function, breaking away from reliance on top sales performers. By automating the extraction of BANTC information and scoring sales meetings, they standardized sales, resulting in an improvement in the organization's overall goal achievement rate from 50% to 90%. The number of orders for members who were struggling increased by up to three times, achieving an overall improvement in the skills of the sales organization.
💡 Success Factors
By establishing a loop of AI-driven meeting analysis, evaluation, and feedback, they created an environment where individuals can autonomously improve their skills. This is a prime example of successfully democratizing sales know-how that was previously dependent on individuals.
3️⃣ Dentsu Digital: Google Cloud x Salesforce Multi-Agent Collaboration Platform
Source URL:
🔗 Summary
Dentsu Digital has launched a multi-agent implementation support service that links "Vertex AI Agent Engine (Google Cloud)" and "Agentforce (Salesforce)." Multiple AI agents share roles and work cooperatively, utilizing integrated CRM data, inventory information, and web behavior logs. This eliminates data silos and consistently supports decision-making across the entire company. It is attracting attention as the first full-scale multi-agent platform support in Japan.
🎯 Use Case Example
An e-commerce customer searches for a "winter coat" -> The Salesforce agent recognizes them as a "red-loving VIP member" -> The Google Cloud agent detects from inventory that "only a few red coats remain" -> Both agents collaborate to autonomously execute a highly personalized proposal.
4️⃣ Kawada Industries x KENCOPA: Construction-Specific Process AI Agent
Source URL: https://prtimes.jp/main/html/rd/p/000000011.000166783.html
🏗️ Summary
Kawada Industries, with over 100 years of history, has introduced the "Kencopa Process AI Agent (beta version)." It automatically generates a draft schedule from design documents and visualizes the AI reasoning process, enabling the confirmation of grounds for schedule review. The goal is to save labor during the estimation phase and facilitate technology transfer. A long-established company has taken the step of introducing a beta version, promoting the formalization of on-site knowledge and moving away from individual-dependent tasks.
🔧 Technical Uniqueness
A specialized agent that understands terminology and processes unique to the construction industry. It systematizes industry-specific tacit knowledge that is difficult for general-purpose LLMs to handle.
5️⃣ LayerX: Attendance Management AI Agent that Reads Work Rules
Source URL: https://prtimes.jp/main/html/rd/p/000000560.000036528.html
📋 Summary
LayerX Inc. will begin offering "AI Attendance Initial Setup," a new feature for its "Bakuraku AI Agent," starting December 23. The AI reads a company's employment regulation PDF and automatically generates complex configuration proposals, such as paid leave accrual rules. This transforms the staff's workflow from "inputting from scratch" to "reviewing AI-proposed settings," fully automating the SaaS onboarding process.
⚠️ Governance Considerations
Final visual verification by the person in charge is mandatory. Emphasis is placed on the balance between AI automation and human judgment.
6️⃣ Continued Support for Building Manufacturing-Specialized AI Agents
Source URL:
🏭 Summary
CEC has begun providing support for building AI agents specialized for the manufacturing industry. By training them on equipment logs, production management data, and blueprints, they inherit the tacit knowledge of veteran engineers. "LegalTech's "MyTokkyo.Ai" reduced document creation time by 60% in the development of human-collaborative robot arms, transforming patent strategy itself. AI is evolving from a "search tool" to an "R&D advisor."
🎓 Accelerating Technology Transfer
To address the aging engineer problem facing Japan's manufacturing industry, AI agents are supporting technical guidance for young workers and predictive maintenance.
7️⃣ The Pinnacle of Vertical AI: Metareal's Industry-Specialized AI Suite
Source URL:
🎯 Summary
Metareal Inc. (Rosetta) has deployed over 50 types of industry-specialized AIs under the "Shigoto-Owaru AI" series. The **Sportswear Demand Forecasting AI (Metareal SW)** analyzes social media, reviews, and marathon schedules to update regional demand scores hourly. The **Pharmacy Supplement Introduction Needs Forecasting AI (Metareal NT)** streamlines sales efficiency through shelf capacity estimation and SKU recommendations. It pinpoints micro-markets that general-purpose AI cannot reach.
📡 Strategic Implications
Orchestrating multiple LLMs with proprietary "Metareal Agents" technology. By mass-producing agents embedded with specific business logic in a short period, they are capturing the long tail of the AI market.
8️⃣ Expansion of New Domains: PaaS-based CMS, Fashion AI, Influencer AI, etc.
Source URL:
🚀 Summary
LeafWorks has launched the Palette CMS Development Support AI (Beta version). Multiple agents act as a team to handle everything from requirements definition to construction. Authentic AI x NEWROPE has integrated 48 trend data AI agents into the fashion AI "Maison AI," reducing market research time from days to minutes. UUUM x FreakOut have jointly developed "GOAT," an AI for supporting influencer marketing video planning, and have begun early implementation.
💼 Acceleration of Horizontal Expansion
The practical application of AI agents is proceeding simultaneously across diverse business domains, including CMS construction, fashion planning, and video marketing.
9️⃣ The Importance of AI Governance and Risk Management Emerges
Source URL: https://prtimes.jp/main/html/rd/p/000000018.000104416.html
🛡️ Summary
Hitachi Consulting has launched "Trusted AI Agent Utilization Consulting." As the autonomy of AI agents increases, risks from hallucinations (incorrect output) and unexpected behavior are growing. They support governed AI implementation, catering to compliance-focused companies and financial institutions. Uzabase manages approximately 50 MCP servers with Kong AI Gateway, achieving both reduced development load and guaranteed security.
⚖️ Problem Awareness
The three major risks identified by Hitachi: 1) Misjudgment, 2) Ethical issues, 3) Data leakage. The division of responsibility and risk management for AI agents are the keys to the next stage.
Market Trend Summary: Three Strategic Shifts
📊 ① Evolution toward "Delegation" and "Autonomy"
As seen in LayerX's work rule settings and Palette CMS's site construction, AI has moved beyond the stage of "supporting" humans to establishing a position as a "delegate" that completes tasks. This signifies the practical supply of digital labor (virtual workforce).
🔗 ② Formation of a Multi-Agent Ecosystem
As demonstrated by the Osaka Prefecture consortium and Dentsu Digital's initiatives, the key to solving social issues is not a single AI model, but an ecosystem where multiple AIs, clouds, and companies collaborate. API-level integration is evolving into inter-agent collaboration, advancing value chain optimization that transcends corporate boundaries.
🎯 ③ Deepening in the Vertical Direction
As shown by the Metareal case, specialized agents are penetrating industry-specific issues that general-purpose LLMs cannot reach. A world view is expected where countless "craftsman AIs" optimized for fields like agriculture, medicine, and logistics are born and interact with each other.
🔮 Outlook for 2026
The succession of major announcements from both the public and private sectors at this timing at the end of 2025 is a harbinger that 2026 will be the "first year of practical AI agent application." In the two years leading up to Osaka Prefecture's 2027 implementation goal, technical issues (security, accuracy, division of responsibility) will be intensively identified and solutions explored. For companies, I feel that data readiness (preparing data that AI can read and write) and the redefinition of organizational culture (BPR) to accept AI agents as "colleagues" will become the source of competitiveness.
いいなと思ったら応援しよう！
この記事は noteマネー にピックアップされました
