In the era of AI, the value of data is being rediscovered.

The 2026 Inclusion · Bund Conference will be held at the Shanghai World Expo Park from September 9th to 12th. Compared to the past focus on 'what can AI do', a more fundamental question was repeatedly discussed at this year's conference: when AI truly enters the industry, how can the massive data in the hands of enterprises be truly utilized by AI?

At the insights forum on "Digital Intelligence Evolution: Making Multimodal Data the Core Fuel of AI", experts and practitioners from universities, research institutions, as well as finance, government, and technology industries engaged in discussions. From banking, insurance, aviation, to agents, digital life, and healthcare, a series of industry practices are providing a common answer: for AI to truly enter the industry, the first step is to let data run into AI.

AI enters the industry, data must first 'run'

Whoever controls the data will gain an advantage in the AI competition. ”

At the forum, Zhou Aoying, former vice president and professor of East China Normal University, pointed out the new position of data in the AI era with this sentence. In his view, data has become an important factor of production, and high-quality AI decision-making cannot be achieved without a high-quality data foundation. "Without high-quality underlying data, any 'ontology modeling' is like a castle on the beach.

Today's enterprises are not lacking in data, the real challenge is how to make data understood and used by AI.

IDC China Enterprise Software Market Research Manager Wang Nan stated that AI is entering the second half, and data capabilities have become the core bottleneck for the implementation of enterprise AI. IDC predicts that by 2035, the size of China's AI data infrastructure market will reach $154.83 billion, an increase of more than 10 times compared to 2025. Meanwhile, by 2025, approximately 90% of newly generated data assets by enterprises will be unstructured data, with a large amount of data scattered across different systems such as NAS and object storage.

The data is increasing, but the connections between the data are not synchronized. This means that a large number of enterprises are holding "rich mines" but do not have the corresponding "smelting capacity".

Wang Nan stated that the future lies in who can first form a complete closed loop covering "data management - data intelligence - data services - agent consumption". The route is to embrace the "small model+multimodal base" model, focus on building a powerful multimodal data base, combine industry knowledge and scenario data, and achieve efficient and controllable production level AI landing through small model fine-tuning.

OceanBase's answer: From multimodal to agent friendly, build an integrated data platform

To truly enter AI, data must first face a reality: the industry itself is multimodal.

An insurance claim may include policy fields, textual descriptions, accident photos, road information, and weather data simultaneously; A flight delay is not just a flight number, but the result of multiple information factors such as crew, weather, air traffic control, and aircraft status working together.

Therefore, the significance of multimodality is not just about "storing more data", but about enabling different forms of data to be managed, associated, and called upon uniformly, allowing AI to see a complete industry.

This is also the starting point for OceanBase to rethink its data base in the AI era. According to Weng Rui, Senior Vice President of OceanBase, the data base in the AI era needs to evolve towards integration and multimodality, allowing different types of data and computing to converge on a unified base, enabling structured, semi-structured, unstructured, and multimodal data to be uniformly governed and interconnected for use.

As AI enters real-world scenarios, the users of data are also changing - agents are becoming new data users.

Wei Peng, the technical leader of Ant Group, shared on site that OceanBase has penetrated the core data link of Ant Group from evaluation, memory construction to online services, supporting user memory accumulation, historical dialogue understanding, and intelligent analysis. Among them, memory and context sharing can save more than 30% of tokens and reduce conversation costs by 60% to 80%.

From people searching for data, to agents using data, and to data becoming the foundation for the continuous evolution of AI products, data is moving from the resources behind the business to the center of AI workflows.

In June of this year, OceanBase released an AI Lakebase that integrates lake and library functions. It natively supports unified storage and processing of structured, semi-structured, unstructured, and multimodal data, allowing AI to truly "understand" enterprises.

Weng Rui frankly stated that in the past, customers wanted a "database that could store data", but now, customers want an "engine that understands data, uses data, and drives intelligence". Based on this change, OceanBase is extending from a database to an AI data platform, accelerating the large-scale implementation of AI in various industries.

Industry practice: OceanBase AI data platform allows industries to see AI

From multimodal data processing to agent support, OceanBase AI data platform is landing in finance, insurance, aviation, Internet and other industries.

At a large state-owned bank, OceanBase improved the accuracy of intelligent question counting from 70% to 98% by introducing a semantic layer to enable the system to understand business intentions.

At a super large property and casualty insurance company, OceanBase LakeBase uniformly associates liability determination, claims, insurance, road network, weather, traffic flow and other data to analyze high-risk intersections and causes of vehicle accidents, promoting the synchronous reduction of accident frequency, claims cost and comprehensive cost rate of car insurance.

In the aviation field, LakeBase manages multiple sources of data including flights, crew, weather, air traffic control, and aircraft sensors, enabling near real-time identification of large-scale delay impact chains, assisting in flight adjustments and crew reassignments, and shortening recovery response times.

At Ping An Property and Casualty Insurance, AI has entered over 1000 business scenarios, with a core scenario AI application coverage rate of 100%, and has accumulated a knowledge base of trillions of tokens in the insurance industry. Next, Ping An Property and Casualty Insurance will build an enterprise level data intelligence platform based on OceanBase's multi-mode integrated architecture and DataSilot.

After data enters AI, the changes are not just about improving business efficiency.

Kuaikan Comics has launched the "Digital Life" Livo based on 13000 IPs, giving characters personality, memory, and growth abilities. OceanBase combines the MaaS platform with LakeBase to achieve unified storage and real-time retrieval of multimodal data, supporting long-term memory and continuous growth of roles.

In the field of medical and health, Wuhou District of Chengdu City and OceanBase announced a "AI+Medical and Health" cooperation, jointly establishing the AI+Medical and Health Data Fusion and Innovation Center, exploring the full chain of medical data collection, governance, circulation, and intelligent application.

From finance, insurance, aviation, to the Internet, health care, different industries are making dispersed data a new driving force that AI can use.

In the view of Yang Bing, CEO of OceanBase, as we enter the era of AI, the focus of enterprise attention is shifting from the performance, cost, and reliability of the database itself to the relationships between different business objects behind the data, as well as their relationships with the real world.

The true commercial value often lies between data that has not yet been connected

When data is reconnected, governed, and organized, and further called upon by AI and agents, data infrastructure begins to move from carrying data to connecting enterprise data with AI.

This is exactly the space that AI data platforms are opening up - connecting the entire data of enterprises downwards, supporting AI applications and agents upwards, and enabling data to move from being "stored and managed well" to being "usable and generating value".

This is also the foundation for the AI new economy to move deeper into the industry. To make the industry see AI, we must first make AI see the industry.