MWC25: NEC CTO's Speaks about Agentic AI and Future Networks [NEC Official]

Yes, you can do that, that's fine. It's always allowed. Yeah, I wanted to be here again at MWC 2025. Actually last year I did a presentation here. My name is Motoo Nishihara and I am the CTO of NEC Corporation. NEC is a global technology company founded over 100 years ago. Our origins are in the telecommunication industry and we also have deep knowledge and expertise in AI technology. Today I'd like to share with you our view of what will be happening with AI and the new requirements for networking. As you know, a lot of companies are talking about agentic AI. We see the concept of agentic AI as key to how AI will be used in the future. NEC just released our own NEC AI agent technology in January, last month, no two months ago. At the core of agentic AI is NEC generative AI technology which we call NEC Kotomi. Since it was released in 2023, NEC Kotomi has achieved accuracy levels comparable to world-leading LLMs using a very small number of parameters. It delivers twice the speed of typical commercial LLMs and is six times faster than open source models. Last year at MWC, as I said, I talked about the concept of an AI orchestrator, which is the integration of many types of AI to solve a specific problem. Our agentic AI is based on AI orchestration. So our agentic AI automatically performs a series of tasks after the user defines the objective. The agentic AI automatically breaks down the objective into several tasks, then it creates a plan for each task, executes those plans, and generates output. What is the difference between the ordinary IT system and the agentic AI? Traditional IT systems are configured based on defined specifications and they perform them exactly. Their behavior is static and the value delivered does not change. However, agentic AI will dramatically transform IT systems from static to dynamic. When business objectives change, agentic AI proactively defines the new required tasks and even generates software to achieve those tasks. It is very, very dynamic. This is the evolution of agentic AI in the future. Today people think of agentic AI as just a tool, but in the near future agentic AI will become our partners and then eventually they will become economic actors. We'll see a time when people and AI coexist and agentic AI companies will emerge everywhere. And individual AI will take a more important role in business-to-business negotiations. One of the key technologies that will enable this is automatic AI negotiation, and NEC is the leader in this technology. Let me show two examples of automatic AI negotiation. The first example is competitive bids by agentic AI in supply chain management. Here is a quick video to show what I mean. The AI representing the purchaser requests 1,000 chipsets for its personal computer production from suppliers B, C, and D. And you can see the suppliers' agentic AI understand the complicated requirements such as delivery times, cost, and even CO2 emissions, and they start to negotiate with their suppliers. Each supplier AI agent reviews the different options and makes recommendations that maximize their own profit. Individual agentic AI can negotiate even when there are conflicts of interest between the different parties. Finally, the purchaser agent AI selects Supplier C in this case as the best option. In all cases, people define negotiation policies and priorities for each agentic AI, then agentic AI can automatically carry out economic activities based on those policies and priorities. Agentic AI will not only work in the cyber world, they will also play a significant role in the physical world. Consider drone traffic management. Today flight plans for drones, especially in Japan, are managed manually. But in the future with thousands of drones, it will be almost impossible for people to manage that coordination. In a proof of concept in Japan using NEC automatic negotiation AI, individual drones with different objectives coordinate with each other to ensure safe operation. This technology could also be used to operate automated vehicles in the future. In the new AI era, agentic AI will communicate with each other and with people to create economic value. In this world, we will need to redesign our networks to connect AI reliably and securely. To take full advantage of agentic AI, we need new network requirements: adaptive and autonomous communication, certification of AI agents, and a transaction of negotiation among AIs based on requirements in each industry. Let's start with communication functionality. In order for AI to act autonomously in the physical world, more sophisticated network control is needed. In some physical world applications such as drones or mobility, random and unpredictable things sometimes happen. Control based on future prediction, such as predicting half a second ahead, will be very important. The operation and management of communication systems will need to be optimized automatically, and automation should include real-time bandwidth sharing to use bandwidth effectively. And NEC's automatic negotiation AI can be used for this. Secondly, agentic AI, because it is digital, can be easily replicated. So think about the risk that malicious agentic AI could easily spread. People could accidentally interact with malicious agentic AI. This is a huge security threat. Then to address these challenges, it is necessary to design and develop AI certification technology and establish a system of certification in our network. Finally, in the application layer, agentic AI will require different algorithms for transactions based on different objectives, such as procurement negotiations like supply chain or automated mobility. Practices differ from industry to industry and there will be a need for industry-specific guidelines and rules. In order to redesign networks, we need multiple perspectives, including not only telecom operators but also application developers and service providers. And NEC is bringing our expertise in AI, communications, and other cutting-edge technology to this type of discussion to accelerate innovation in this field. NEC is leading various initiatives such as launching public competitions and consortiums. Our work has been recognized by the Scientific Council of Japan as a part of the future academic advancement initiative. The council is supported by the Japanese government and represents the Japanese scientific community around the world. So lastly, I'd like to raise the issue of the role of people in an age where we coexist with AI. To address this question, I'd like to suggest a concept that comes from Japan 400 years ago. The concept is called Sampo Yoshi, which means three-way good. So this word Sampo Yoshi means that good business benefits the seller, the buyer, and also society. This idea is widely adopted as a business policy in Japan and in NEC. In the future where people and AI coexist, this idea will continue to be more important. And people's role will be to ensure that the agentic AI always considers the society perspective along with business objectives like profit. Currently, the way to deal with AI risks such as deepfakes and hallucination is just to regulate AI behavior. But regulation alone cannot cover every situation. We need to build a more fundamental trust mechanism within agentic AI. What we propose is that people set the Sampo Yoshi concept as a foundation for the behavior of agentic AI and allow agentic AI to freely create economic value based on that foundation. Then we will see new kinds of economic value consecutively emerge, as long as we ensure that AI is always helping society. The spirit of Sampo Yoshi is aligned with NEC's purpose. We believe that technology enables a more sustainable world and provides the opportunity for everyone to reach their full potential. With this purpose in mind, we are committed to working with industry leaders and stakeholders as we enter this new AI era. That's all my presentation. Thank you.