Creating Value Through Big Data, IoT and AI

「NECは、災害時に自治体などへ集まる現場画像から、画像内に小さく写る被害を高精度に把握するVLM(Vision-Language Model、視覚言語モデル)局所領域理解技術(注1)と、VLMの認識能力を維持したまま新たな被災事象を追加できるVLM継続学習技術(注2)の2つを新たに開発しました。」

「Jointly developing safe and reliable business-specific AI solutions」

「We developed an AI agent that diagnoses system security risks and an AI agent that supports internal audits related to information security, using our proprietary generative AI "cotomi."」

「We developed “cotomi Act” (Note), a business-automation AI agent technology that achieved the world’s first task success rate exceeding that of humans, and began providing solutions combining “cotomi Act” with software, consulting and operations and maintenance services in January 2026.」

「We developed agentic AI (Agentic AI/AI Agent) that enables the automation of specialized tasks using our generative AI, cotomi, and began offering it as “NEC AI Agent” in January 2025.」

「We developed FireDucks, software that accelerates Python-based data analysis by using all cores of a multicore CPU to process large volumes of data in parallel and efficiently, increasing the speed of data preprocessing required for data analysis by up to 16 times, or approximately 5 times on average, compared with conventional methods (IT Services business). ・As technology supporting more efficient operations and innovation in business processes through video, we developed technology combining an LLM with video-recognition AI to automatically generate shortened videos and explanatory text tailored to users’ objectives from long videos (IT Services business)」

「Using proprietary innovations and technologies, we developed an LLM with high Japanese-language performance that is lightweight and fast and can operate in both cloud and on-premises environments, while limiting its number of parameters to approximately one-thirteenth that of leading large language models overseas (Large Language Model (LLM)) (IT Services business).」

「We developed technology that remotely analyzes bridge conditions using AI trained on displacement data obtained through satellite-radar remote sensing, bridge structures and temperature changes. By accurately detecting abnormal bridge deflection, which was previously difficult to identify, the technology enables the early detection of serious damage that could lead to bridge collapse.」

「We developed technology that recognizes a wide variety of features in different formats from camera footage at work sites, including people’s postures, the types of tools they are using, locations and surrounding environments, and analyzes their relationships in an integrated manner using deep-learning methods. The technology enables highly accurate, real-time recognition of diverse work activities even in environments where many people are moving around, contributing to the digitalization of work-site activities, work-style reform and higher productivity in workplaces facing labor shortages (Enterprise business).」

「・To help create safe and secure working environments at construction sites and factories using artificial intelligence (AI), we advanced “weakly supervised learning” technology, which can use ambiguous information in the data learned by AI. By overcoming learning instability caused by ambiguous information, we developed technology that significantly simplifies the creation of training data required for AI-based image recognition and object detection (Other) ・As part of efforts to promote the digital transformation (DX) of infrastructure facility management, we developed technology that remotely analyzes bridge conditions using AI trained on displacement data obtained through satellite-radar remote sensing, bridge structures and temperature changes. By detecting abnormal bridge deflection, which was previously difficult to identify, with millimeter-level accuracy, the technology enables the early detection of serious damage that could lead to bridge collapse (Social Infrastructure business)」

「To help create safe and secure working environments at construction sites and factories using artificial intelligence (AI), we advanced “weakly supervised learning” technology, which can use ambiguous information in the data learned by AI. By overcoming learning instability caused by ambiguous information, we developed technology that significantly simplifies the creation of training data required for AI-based image recognition and object detection (Other).」

「We worked to further strengthen the NEC Digital Platform.」

「As part of efforts to realize AI that supports advanced decision-making in collaboration with people, we developed AI technology that applies insights obtained from analyzing brain activity during complex decision-making. By acquiring and analyzing data simultaneously, the technology is expected to accelerate the processing of facial recognition and cyberattack detection and analysis by up to 20 times while maintaining high accuracy (Global business).」

「As part of efforts to realize AI that supports advanced decision-making in collaboration with people, we developed AI technology that applies insights obtained from analyzing brain activity during complex decision-making. By acquiring and analyzing data simultaneously, the technology is expected to accelerate the processing of facial recognition and cyberattack detection and analysis by up to 20 times while maintaining high accuracy (Global business).」

「We developed “goal-oriented task-planning technology,” an AI technology that automatically sets robot work procedures and determines and controls the most efficient and safe movements simply by having an on-site worker specify the work objective, without requiring advance procedure design by a robotics specialist (Enterprise business). We also built an operations support system combining “logical thinking AI,” an AI technology supporting the stable operation of large-scale infrastructure such as chemical plants, with a mirror plant reproduced in a simulator. We demonstrated that the system could achieve operations 40% more efficiently than manual operation by an operator (Enterprise business).」

「・To help prevent the spread of COVID-19, we developed social-distancing assessment technology that analyzes camera footage from public facilities such as stations and airports and stores and visualizes crowd density in real time without identifying individuals (Global business) ・As part of efforts to support COVID-19 vaccine development, we applied AI prediction technology developed through work on personalized cancer immunotherapy to analyze the COVID-19 genome and published the results of the analysis (Other)」

「We developed an “inter-AI automated negotiation platform” that coordinates multiple AI systems involved in intercompany transactions across the supply chain, from processing and assembly through logistics, enabling rapid and effective coordination among those AI systems.」

「By enhancing and applying “rare-event detection technology,” which enables AI to repeatedly run simulations while learning and efficiently identify, in a short time, defects that are difficult to find during the design stage because their probability of occurrence is extremely low, we demonstrated that the time required for evaluation can be reduced by efficiently identifying defect patterns that even experts would have difficulty anticipating in the preliminary evaluation of production processes involving multiple product types.」

「Applications of AI included verification with medical corporation KNI of early detection of signs of changes in patients’ conditions toward realizing digital hospitals, and the launch of a “supply-demand optimization platform” that optimizes supply-demand balances across the entire value chain to address the social issue of food loss and waste.」

「We developed “rare-event detection technology,” which enables AI to repeatedly run simulations while learning and efficiently identify, in a short time, multiple defects that are difficult to find during product design because their probability of occurrence is extremely low. ・We developed “time-series data model-free analysis technology,” which uses data collected and accumulated from sensors and other devices installed in social infrastructure systems such as plants to rapidly and accurately determine system conditions, detect abnormalities, diagnose failures and predict breakdowns.」

「We developed “rare-event detection technology,” which enables AI to repeatedly run simulations while learning and efficiently identify, in a short time, multiple defects that are difficult to find during product design because their probability of occurrence is extremely low.」

「To address these challenges, we developed data-processing technology for statistical machine learning suited to vector computers. …It enables processing at speeds more than 50 times faster than conventional general machine-learning methods.」

「・We developed technology for remotely, accurately and cooperatively controlling multiple robots in real time via a wireless network (Enterprise business) (Note) This development was carried out jointly with Nidec Corporation. ・We developed automated optimization technology for deep learning that facilitates improvements in AI recognition accuracy (Public business)」

「・We developed data-processing technology for statistical machine learning, a form of AI used to analyze and predict outcomes from large datasets, expanding the application areas of vector computers (System Platforms business) ・We developed “identification dot technology,” which enables a 1 mm handwritten dot to be used as an identification tag for miniature electronic components where a barcode cannot be attached (Enterprise business)」

「We developed “multimodal image fusion technology,” which uses AI to automatically and effectively combine visible-light images captured by ordinary cameras with invisible-light images captured by thermal cameras and other devices. The technology significantly improves the visibility of objects and conditions that are difficult to identify from either image alone and enables rapid detection of abnormalities and hazardous objects even in poor environments (Public business).」

「We launched the technology brand “NEC the WISE” in the fields of AI (artificial intelligence) and IoT (Internet of Things), which are key to differentiating our Social Solutions business, and promoted the development and use of related technologies.」

「We developed two methods for confidential computing technology, which processes data while it remains encrypted and strongly prevents the leakage of confidential information: a “basic algorithm that dramatically improves processing speed” and a “high-speed search method enabling aggregation in databases” (Public business, Enterprise business). We also developed “automated predictive analytics technology,” which enables even non-specialists to perform fast, accurate predictive analysis of vast volumes of data stored in relational databases and can be widely used in business systems (Enterprise business).」

「We developed “automated system update AI technology,” which rapidly updates ICT systems without interrupting services in systems providing socially important, non-stop services (Public business). We also developed “customer profile estimation technology,” which automatically and accurately estimates detailed profiles for each customer, including occupation, preferences and annual income, based on relatively accessible basic profiles such as age and gender and purchase histories, without the involvement of marketing specialists (Enterprise business).」

「We developed network technology that remotely and centrally controls IoT (Internet of Things) devices, such as those used on factory production floors, in real time over a wireless environment, enabling flexible changes to production lines (Enterprise business).」

「In big data, we provide systems supporting a range of business operations, including systems that automatically analyze large volumes of text documents containing the “voice of the customer” received daily by financial institutions, while also strengthening initiatives involving IoT (Internet of Things), in which all kinds of objects are connected to the internet, and artificial intelligence (AI).」