Quollio and NEC Complete Technical Verification of Context Layer for Managing AI-Extracted Features and Business Context in a Catalog; Internal NEC Pilot Begins

Quollio Technologies (Headquarters: Minato-ku, Tokyo; CEO: Ryota Matsumoto; hereafter "Quollio"), a provider of next-generation data intelligence solutions for enterprise companies in Japan, has collaborated with NEC Corporation (Headquarters: Minato-ku, Tokyo; President and CEO: Takayuki Morita; hereafter "NEC") to complete technical verification of a system that integrates "Quollio Data Intelligence Cloud" with the AI data analysis platform "dotData" (Note 1). This integration allows business context to be managed and utilized as business metadata for features automatically extracted by AI. Following these results, both companies began a pilot program in NEC's internal business environment in March 2026.

■ Background of Joint Verification: Challenges in AI-Ready Data Preparation

With the spread of generative AI and AI agents, the importance of converting data accumulated within companies into "knowledge (AI-Ready data)" that AI can understand and utilize is increasing. "dotData" streamlines and accelerates the preparation of AI-Ready data by automatically extracting features (insights), which are statistical facts hidden in data. However, "semantic information"—such as how to interpret these extracted features within a business context and translate them into decision-making—often relies on human expertise and operational experience, creating a barrier for AI when generating responses based on context.

■ Content of Technical Verification: Establishing a "Knowledge Circulation Cycle" on a Data Platform

From October to December 2025, the companies utilized "Cortex Agents," a generative AI agent provided by Snowflake Inc., and "Streamlit," an application development tool for visualizing and sharing data. In this environment, through dialogue (brainstorming) via the AI agent's chat UI, it was confirmed that by adding business context to the features extracted by dotData, the AI agent could generate meaningful responses that go beyond simple numerical presentation by considering the context. Furthermore, the technical feasibility of an architecture that autonomously integrates and utilizes these elements was also confirmed.

1. Verification Architecture: Features extracted by dotData were stored in Snowflake, and an AI agent environment was built to call Cortex Agents from Streamlit to interact with users.

2. Fusion of Statistical Facts and Business Context (Concrete Example): Supermarket purchase data was used as a model case for verification. Through dialogue (brainstorming) in the AI agent's chat UI, "business context" was derived for the features extracted by dotData. By utilizing both the features and the business context together, it was confirmed that the AI agent could provide responses beyond mere numbers.

• Fact extracted by dotData (Feature): "Purchase time is in the 22:00 hour."
• Business context derived through chat UI brainstorming: "Customer visiting just before closing (last-minute demand)."
• AI Agent Behavior: When a user asks, "What are the characteristics of customers who buy Product A?", the AI references the "22:00 hour (fact)" data and incorporates the "last-minute customer (business context)" to generate a response that facilitates user understanding...