[ITmedia Enterprise] NEC prepares 1,435 data catalogs in one year, taking on the challenge of decision-making support using AI

Mr. Seki says, ``Increasing the use of data to realize data-driven management and maximizing its value cannot be achieved in one step.We started with data utilization through BI (Business Intelligence), then proceeded step by step to formulate an overall optimal data strategy and create an actual data management organization.The ``One NEC Data Platform'' was put in place as the IT infrastructure to make this initiative possible.''

The One NEC Data platform is a platform that collects and accumulates data scattered in the cloud and on-premises in Japan and overseas in real time, and utilizes data virtualization technology (technology that virtually integrates and handles data from different locations without duplicating or moving data). NEC uses the collected data for dashboard visualization and process mining, and has also created a data catalog for this purpose.

Mr. Seki says, ``In the first year since we started the initiative, we have compiled 1,435 data catalogs and built numerous new dashboards.The dashboards have been used to visualize management, operations, and IT, and everyone from employees to top management has come to use them for decision-making.The fact that we have established the use of data is a major success of our internal DX.'' It took one year to complete the system since there was no data catalog. There are 44 new dashboards. By building this data infrastructure, NEC has established a system to tackle AX.

Expert AI appears at meetings and speaks: Work supported by the generative AI platform “NGS”

NEC aims to become an AI-native company in five years' time by viewing the impact of AI on management in three stages: ``Utilization of AI for individual purposes,'' ``Utilization of AI at the company-wide level,'' and ``Becoming an AI-native company.'' NEC positions companies that go beyond the stage of AI supporting people and create an AI-centered world as AI native companies.

According to Seki, many Japanese companies, including NEC, are at the stage of using AI for individual purposes. Under these circumstances, NEC is moving to a stage where its organization and business processes themselves are being transformed based on AI. "The source of competitiveness is to make full use of AI to improve the productivity of people and organizations, change the way time is used, and create new corporate value. Companies that cannot put this into practice face the risk of being eliminated in five years. We are working with a sense of crisis that we too will be eliminated." (Mr. Seki)

In order to become an AI-native company, NEC has developed an in-house generative AI platform called "NEC Generative AI Service" (NGS), which incorporates various AI functions to help reform business processes. "Our company has been researching and developing AI for the past 50 years. We are incorporating these technologies into NGS and making full use of them. Global strategic collaboration is also important. By combining our technologies with those of our partners, we can achieve things that cannot be achieved with our own technology alone, and we will pass this on to our customers as Client Zero." (Mr. Seki)

At NGS, employees can select and use multiple LLMs (Large-Scale Language Models), as well as RAG (Retrieval Augmentation Generation) and Fine-Tuning models that incorporate company-specific information, making it possible to safely utilize AI by combining LLM and company-specific information. "Knowledge AI" has been prepared as a common application layer that can be used for a variety of tasks. In the upper layer, individual applications are built for each industry and business area, and AI and AI agents are incorporated.

Knowledge AI is a specialized AI that employees can create by simply uploading files, activities, minutes, comments, etc. There are two types of AI: ``alternate AI'' that is based on the thinking and values ​​of chief executives (CxOs) in various fields, such as CEOs and CIOs (chief information officers), and ``expert AI'' that has specific specialized knowledge. In addition to communication such as reviews, expert AI performs tasks that require expertise, such as research analysis, document creation, and answering questions, on behalf of humans. In the six months since its release, more than 2,000 Knowledge AIs have been created.

Mr. Seki says, ``When using AI from a management perspective, we are improving the quality of management by reviewing documents and having conversations with AI.Also, specialized AI participates in 30,000 meetings a day and speaks, reviews materials, and gives advice, contributing to improving the quality of meetings, lively discussions, and improving the accuracy of decision-making.Furthermore, after an action is planned, we are also working to improve the speed of execution by tracing the execution, detecting risks, and providing advice.''

How can we prepare data that can be used by AI?

"We have developed a data catalog because humans need a data catalog to utilize data. However, when it comes to AI utilizing data, we have found that a mere data catalog does not advance data utilization or AI-based decision-making." (Mr. Seki)

The management cockpit (a group of dashboards that visualize and analyze important company-wide indicators), which is the center of NEC's data-driven management, has been built over the past five years and incorporates AI to support decision-making. Decision-making support using AI is currently progressing in areas such as business management, marketing, and auditing. Previously, AI utilization was limited to structured data, but we are currently working on AI utilization that utilizes vast amounts of unstructured data such as email, chat, and voice data. Mr. Seki says, ``The NEC Group has over 30PB (petabytes) of unstructured data, and we are planning a system that can use this data to have conversations with AI.'' According to NEC, this mechanism has already been realized as of the end of October 2026.

At the time of the presentation, NEC was proceeding with the development of incorporating NGS technology into Snowflake Intelligence, an AI agent provided by Snowflake that interacts with data using natural language (ed. note 1). The aim is to strengthen the ability to understand documents and derive suggestions from analysis results. The ability to deeply understand the meaning of not only text data in a document but also diagrams and tables is made possible by NEC's diagram context understanding technology. Additionally, Knowledge AI's expert AI enhances analysis and suggestion presentation, and incorporates the ability to create reports with deeper insights.

Through the use of Snowflake Intelligence, the importance of "understanding the meaning" of data for AI has become clear, so we are also promoting the strengthening and expansion of the semantic layer (the layer that collectively manages the meaning of data and business definitions). As a core function of the One NEC Data platform, we have a system called the "AI Ready Platform" that incorporates a layer of semantic ontology (a system that defines the meaning and relationships of data in a form that can be interpreted by AI) that gives meaning to data.

"The AI Ready platform has developed ontologies related to management, processes, work styles, IT security, etc. While also utilizing ontologies from other companies, we have also developed NEC's own ontologies and integrated them to create NEC's domain ontology. This ontology is in a form that allows AI to interpret the meaning, and the input is data catalogs and data sources that have been maintained over the past five years." (Mr. Seki)

Seki explains that ontologies are typically created by engineers called FDEs (Forward Deployed Engineers) who read data. FDE refers to engineers who go into customer sites and work on data integration and system construction. NEC is developing a function called ``AI FDE'' to automate this work. We are currently at the stage where the AI ​​itself is attempting to read data catalogs and data sources and create a domain ontology overnight. In order to create AI for decision-making using this ontology, it is necessary to develop an AI agent that utilizes the ontology. NEC is already using AI to develop AI agents, and plans to switch to AI-driven development in the future as well.

Mr. Seki says, ``With the AI ​​Ready platform, we aim to improve NEC's own data utilization and decision-making.''

The data catalogs that have been prepared for human use alone cannot become data that can be used by AI. NEC is using this as a material to create an ontology that can be interpreted by AI, and is also leaving the creation of the ontology itself to AI. We have entered the stage of adding layers of "meaning" for AI to the data infrastructure that we have spent five years building.