Happy New Year! This is Miki Hirano from Cinnamon AI.
Thank you very much for your generous support and interest in Cinnamon AI's efforts last year.
Thank you for your continued support this year.

In 2026, AI will move from the experimental stage to the full-scale value creation phase.

In the global economy and geopolitics, the US-China conflict is polarizing technology and investment, while each country aims to strengthen its own advantage and governance.national AI strategyWe are promoting the following.

Companies are moving beyond PoC andIt's not about "what you can do" but "where to use it"I started to focus.

On the technology side, multimodal AI and agent-based AI are entering a period of maturity. In major industries, leading companies are achieving great results by "redesigning the business itself, rather than automating partial optimization," and we are beginning to see the turning point at which AI utilization will lead to full-scale implementation.

On the social front, the impact on employment and skills is becoming a reality, and ensuring trust in AI, including accountability and data governance, will be key to competitiveness.

In the investment environment, AI-related investments now account for the majority of VC funding, and collaborations and acquisitions between platform companies and startups are becoming more active.

Below are some of the trends in AI that are evolving at such incredible speed.

AI-first operations

Below is a look at what happened in 2025 and the predicted technological advances for 2026.

technology	What happened in 2025	2026 Predictions
LLM	Its use in business is becoming commonplace. Copilot-type (AI used by humans) is the mainstream. The latest LLM shows an IQ exceeding 150, making it more intelligent than almost all humans.	From standalone useBusiness process integrationInstead of choosing a general-purpose LLM, the evaluation criteria are stability, cost, and operability rather than performance.
RAG	Rapidly spreading as a countermeasure against hallucinations. Production use has begun for internal document search and QA. Many PoCs and partial operations.	Models themselves continue to learnRAG is a place for self-improvement, constantly updating its knowledge.Rather than being a standalone RAG, it is integrated with AI agents and business flows.
AI Agents	PoCs have increased rapidly. General-purpose AI agents initially attracted attention, but agent workflows have become mainstream. Limited autonomous execution has begun for some tasks, but there are many cases where it has been stopped due to unclear value and concerns about control.	Only agents that work within the business will survive.Specialized business model with limited objectives and authorityThis trend continues, and it is at the core of AI-first operations.
Multimodal	Integrated understanding of text, images, and voice has reached a practical level. It has begun to be used for UI/UX improvement and on-site support.	Standard interface. AgentEyes, ears, handsIt will be incorporated as a system to accelerate the use of AI in manufacturing, medical care, logistics, and other fields.

Although such technology is evolving every day, AI is still not omnipotent. In 2026, rather than replacing human work,Redesigning processes for AIThis will be required.

The more agents there are, the more likely it is that problems will occur due to malfunctions, miscommunication, incorrect permissions, and inadequate logging rather than hallucinations.

Therefore, it will become necessary to redefine operations based on the premise of where using AI will have the greatest impact.

For example, Cinnamon AI is turning appraisal work in the financial industry into AI agents, but it is not using AI evenly. Of the multiple appraisal processes, the earlier areas such as the first and second appraisals are shifted to AI, and the latter areas are left to human work.

Also, what's important in AI-first Ops isOperational quality over intelligence(Reliability).

Reliable AI (which will not break during operation) will be made into a field requirement, including explainable AI.

In this process, people's roles will change from workers to exception handlers, decision makers, and quality managers.

Cinnamon AI has a variety of examples of AI agents and RAGs, so if you're interested,Inquiriesplease.

AI for Science

Research → Hypothesis → Experiment → DocumentationAutomate and speed up the process, and create new products from that cycle.discoveryAI for Science maximizes the probability of success.

Phase	maturity	Current status and characteristics
investigation	★★★★★	It can search and summarize literature, experiments, and observation data across the board. It is already being implemented in practical and research settings.
hypothesis	★★★☆☆	It is possible to present high-dimensional, non-intuitive hypothesis candidates. It is effective as a search support, but the selection is human-dependent.
experiment	★☆☆☆☆	In cutting-edge cases, automated laboratories and control are on the rise. Although progress is being made due to the physical implementation involved, generalization is slow.
Documentation	★★☆☆☆	Although it is possible to generate text, human judgment is essential for descriptions that require verifiability, reproducibility, and scientific responsibility.

2025 saw a major breakthrough in research automation thanks to Deep Research and RAG.

In 2026, AI will go one step further than just "suggesting candidates" and will likely see an increase in cases where automated experiments (sample preparation → synthesis → measurement → analysis) are connected. The bottleneck will be experimental standardization, measurement data quality, and safety, rather than model accuracy.

Japan is strong in equipment, materials, and manufacturing processes, so standardizing lab automation x equipment x data specifications could be a winning strategy.

Cinnamon AI uses RAGIn-house technical documentation researchThere are some examples of this, so if you are interested,Inquiriesplease.

Physical AI (combination of robots and AI)

Physical AI is a technology that directly interacts with the physical world through actuators such as robotic arms and motors.physicalityIt is an intelligence that hasautonomouslyTake action.

Conventional robots simply repeat the instructions given by humans, but Physical AI will evolve into robots that can autonomously change and evolve their behavior depending on the situation.

▽ Physical AI for next-generation manufacturing
Improving productivity and automating manufacturing, a specialty of Japan, is a top priority. In particular, Japan should take the lead in areas such as introducing autonomous robots to automobile and electronics production lines, applying AI to precision assembly robots, and replacing skilled workers with AI robots.

▽ Physical AI for disaster prevention
Japan is a country prone to natural disasters such as earthquakes and typhoons, so disaster prevention technology is directly linked to the lives of its citizens. As disasters increase due to climate change, developing disaster response robots is a mission that Japan should lead the world on.

In 2026, the convergence of robotics infrastructure models, physical simulations, and edge inference will likely lead to an increase in practical solutions for implementation in warehouses, factories, infrastructure, and more.

The leading implementation area isIndoor/semi-structured environmentThis is because environmental fluctuations are limited and KPIs (productivity, safety, labor shortages) are clear.

• Warehouse/Logistics (Picking assistance, transportation, sorting)
• Factory (inspection, setup support, material supply)
• Infrastructure maintenance (patrols, inspections, assistance with dangerous work)
• Service (customer service and guidance in limited environments)
  

The realization of humanoid robots is a dream, but we will first have to focus on winning strategies for each task category. I think that design for unit tasks, such as repetitive tasks like lifting, carrying, and stacking, or tasks in dangerous areas, will become mainstream.

AI as a national strategy

A major change in 2025 is that the United States and China formulated their national AI strategies in July, the EU in October, and Japan in December.

AI not only increases productivity, but also allows the technology to be used for military, economic, information, and diplomatic purposes, which is directly linked to national power.

Japan has also recently been undergoing major changes due to the rise of AI, semiconductors, batteries, energy, and economic security. From the postwar period through the 1980s, the country was driven by government-led industrial policy, but over the past 30 years, the country has shifted to a private-sector initiative, so I feel that we are at a turning point.

A comparison of the national AI strategies of the three countries + EU is as follows.

perspective	🇯🇵 Japan (Artificial Intelligence Basic Plan)	🇺🇸 United States (AI Action Plan)	🇨🇳 China (Global AI Governance Action Plan)	🇪🇺 EU (Apply AI & AI in Science)
Basic purpose	AI utilization andSafety and reliabilityand promoting the social implementation of AI	AI Competitiveness・Maintaining leadershipAccelerating innovation	Global Collaboration・Safety, fairness, openness	Accelerating AI adoption,Strengthening industrial competitiveness,scientific research
Key Policies	Promoting utilization, strengthening development capabilities, governance-driven, social change	Prioritizing innovation, deregulation, infrastructure and competitiveness	Opportunity sharing and cooperation, expansion of industrial applications, international cooperation	Apply AI to industry, strengthen AI scientific research (AI in Science)
AI and Security	Emphasis on sovereignty and trust, and development of social infrastructure	Security andCompetitive StrategyCenter of	Secure and manageable	Strengthening technological sovereignty and autonomy
Infrastructure Strategy	Computing resources and data infrastructure development	Data centers/semiconductors/electric power etc.	International cooperation to strengthen digital infrastructure	AI Gigafactory / AI Factory etc.
Focus on industrial use	Various fields including public, medical, manufacturing, and on-site AI	Wide range of industrial and government applications	AI application to all industrial sectors	Manufacturing, medical, energy, defense, etc.
Research and Science Strategy	Basic research, human resource development, and international collaboration	Promoting basic and applied research	Strengthening international cooperation frameworks	Research Resource Integration through the RAISE Initiative
Data Strategy	Promote data utilization and sharing	Promoting development through data infrastructure development	Promoting data openness and sharing	Data Union Concept
Governance/Regulation	Balancing risk response and promotion	Regulations are an impediment to competitivenessTendency to consider	Emphasis on safety, ethics, and fairness	AI Act (risk-based)
International expansion	Involvement in international cooperation and rule-making	Cooperation with allies and international leadership	Emphasis on open international cooperation	International rule-making and cooperation
Use case focus	Addressing social issues and optimizing industry	National defense, government, and general civilian affairs	General industry and social development	Industrial Competitiveness, Science, and Public Affairs
AI for Science	Priority Areas(drug discovery, materials, climate, quantum, computational science) Explicitly positioned in national strategy	Extremely important(National defense, energy, life sciences, space) Source of national competitiveness	emphasis(Basic science + national key projects) Mainly for domestic use	Top priority(AI in Science is an independent strategy) RAISE across Europe
Physical AI	Priority Areas(Manufacturing, logistics, medical care, nursing care, infrastructure, disasters) Directly linked to measures to combat population decline	Most important in military, space and autonomous weaponsThe private sector is left to the market	national emphasis(Robotics, Automation, Smart Manufacturing) Social Control and Industrial Competitiveness	limited(Focus on industrial robots and automation) Be cautious about humanoids, etc.

The three countries plus the EU share a common goal of placing AI at the core of their national strategies, strengthening AI infrastructure, and promoting its implementation in society.

The United States wants to remain the dominant power.
China's capabilities are on par with those of the United States, but it wants to be used by the world.
Japan wants to reduce its dependence and use it as a device to maintain national power rather than a device to expand it.
The EU wants to make rules so it won't be dominated.

The color of the strategy

Japan: Focus on achieving both promotion and safety and reliability
US: Competitiveness, leadership and deregulation at the forefront
China: Emphasis on international cooperation, safety, and fairness
EU: Balancing industrial adoption, scientific research, technological sovereignty, and risk-based regulation

There are differences such as:

As each country sets its national AI strategy, I myself am also working on theJapan Growth Strategy CouncilHe took office in November 2025.

It is also a subordinate organization of the Japan Growth Strategy Council.AI and Semiconductor Working GroupandDrug Discovery and Advanced Medicine Working GroupHe was appointed as an expert in December 2025.

The New Capitalism Realization Council under the Kishida and Ishiba administrations has made many recommendations regarding AI, and now that AI has become a cornerstone of national strategy, we will continue to work hard to contribute even a little to Japan's AI strategy.

Also,In an era where what AI can do is expanding at an incredible speed, determination is importantA book about willpower is scheduled to be published by Toyo Keizai Inc. in March 2026, so please look forward to it.

Thank you for your continued support this year.

Cinnamon Co., Ltd.
President and CEO
Mirai Hirano