PingCAP Partners with MySQL Solution Specialist Pasona Data & Design

PingCAP Co., Ltd. (Headquarters: Chiyoda-ku, Tokyo, President: Eric Han, hereinafter PingCAP) announces the conclusion of a partnership agreement with Pasona Data & Design Inc. (Headquarters: Minato-ku, Tokyo, President: Hajime Kono, hereinafter Pasona Data & Design), a company specializing in database, cloud, and AI technologies.

Background of the Collaboration: The Need for Data Infrastructure to Support the "Intelligence" of AI Agents

As the use of generative AI and AI agents shifts from the "experiment" phase to "practical application," the integration of data from existing core systems with AI has become a severe bottleneck for many companies. For AI agents to autonomously make correct decisions, accurate data and appropriate search methods are required. The complexity of traditional data platforms presents the following barriers to AI implementation:

1. Inability to know the "present" (Lack of real-time capability) When core systems and the data used by AI are separated, a time lag occurs in data synchronization between them. If an AI agent acts based on inventory data from hours ago or outdated customer status, it can lead directly to erroneous decisions and complaints.

2. Low-quality answers due to a single search method For AI agents to infer with high accuracy, they must reference relevant data using methods like RAG (Retrieval-Augmented Generation). Searching diverse corporate data requires not just simple exact-match searches, but also methods like vector search and full-text search. Without these, the AI cannot sufficiently draw upon referenceable knowledge, leading to a degradation in accuracy where it cannot return the answers it originally should.

3. Reduced development speed and increased costs due to the "infrastructure wall" Traditional data platforms do not account for the on-demand execution and frequent schema changes characteristic of AI services. If AI services are integrated under these conditions, development resources are consumed by "investigating impacts on existing infrastructure and accommodating changes" rather than the actual "implementation of AI features."

Against this backdrop, there is a growing demand for a scalable data platform that maintains core system data while supporting multiple search methods. Furthermore, deep understanding of existing data assets becomes crucial in the practical operations of AI agents. Through this partnership, Pasona Data & Design—one of Japan's leading groups of MySQL specialists with over 20 years of support experience in MySQL consulting and over 1,000 corporate implementations domestically—and PingCAP aim to provide a scalable infrastructure environment supporting next-generation AI utilization. This will be achieved using the NewSQL database "TiDB Cloud," equipped with the new architecture "TiDB X" optimized for AI agents.

Solutions Provided by TiDB Cloud / TiDB X

1. Seamless integration through high MySQL compatibility

Leveraging Pasona Data & Design's core strength in MySQL expertise, existing data on-premises or in the cloud can be integrated into TiDB Cloud without requiring major rewrites of application code. This enables the unification of the data infrastructure for both core systems and AI services. Even when considering the construction of a new AI service, utilizing TiDB Cloud with its high MySQL compatibility minimizes learning and operational costs, allowing teams to concentrate solely on AI service development.

2. TiDB X: A new architecture handling the explosive load of AI agents

Unlike humans, AI agents issue a massive volume of complex queries in parallel. To meet these demands, TiDB X is equipped with the following features:

- Serverless Auto-scaling: It instantly expands or contracts the database's computing resources in response to the agent's activity level. This prevents unnecessary costs while ensuring business continuity without service interruptions, even during sudden spikes in load.

- Native Support for Vector Search: Vector data, essential for RAG, can be handled directly via SQL. There is no need to prepare a separate dedicated vector database; "structured data (such as purchase history)" and "unstructured data (such as manuals and conversation logs)" can be combined and searched at ultra-high speeds within a single database.

3. Real-time HTAP (Hybrid Transactional/Analytical Processing)

Because TiDB Cloud can simultaneously perform transaction processing (OLTP) and analytical processing (OLAP) on a single platform, it provides the data analysis crucial for AI services within the same environment. This enables a rapid cycle of improvement based on feedback from the field.

Expected Use Cases for TiDB Cloud Leveraging AI Agents

T