In 2025, AI intelligent agents transitioned from concept to industrial application, yet few players have successfully convinced enterprise clients to purchase and repurchase their solutions. The first annual report since its listing from Haizhi Tech (02706) provides a compelling answer with key figures: annual revenue reached 621 million yuan, a year-on-year increase of 23.4%. Revenue from the Atlas Intelligent Agent solution was 146 million yuan, surging 68.4% year-on-year, with the gross profit margin rising to 53.2%. Adjusted net profit was 24.15 million yuan, up 42.6% year-on-year, indicating sustained profitability.
These numbers reflect the complete logical chain of a B2B technology company: strong technology, a comprehensive product suite, and deep competitive moats.
A Solid Foundation: From Graph Database to Full-Stack Product Matrix Haizhi's competitiveness is fundamentally built on robust underlying technology. In 2025, the company led a major national-level project focused on key technologies and industrialization of a new generation of multi-modal databases, using graph computing to construct a trustworthy and governable underlying computational core for AI execution systems in highly complex business scenarios. That same year, it was selected as the sole lead for a key Beijing municipal project titled "Construction and Demonstration Application of a Joint Graph-Model Application Platform Based on Large-Scale Graph Data Analysis Technology." The new version of its AtlasGraph database saw significant enhancements in dozens of core functions, including temporal graphs and data lifecycle management, markedly improving system stability and processing efficiency.
Building on this technological foundation, Haizhi has established a full-chain product matrix covering enterprise digitalization to intelligent transformation—comprising the DMC data intelligence platform, Atlas knowledge graph platform, AtlasGraph database, and intelligent agent application platform. This creates an end-to-end闭环 from data governance and knowledge modeling to intelligent decision-making and automated execution. Essentially, Haizhi can provide all the necessary products for an enterprise to progress from initiating digital transformation to completing an intelligent upgrade.
This full-stack capability directly translates into strong customer stickiness. By the end of 2025, the company had served over 400 industrial clients across multiple sectors. It has successfully implemented domestic replacements for graph databases in four large state-owned commercial banks and continues to expand upwards from this base, developing graph applications and intelligent agent products. The deeper the foundational technology is rooted, the easier it is to expand the upper layers—this constitutes Haizhi's core moat, distinguishing it from other AI companies.
The True Proposition of Intelligent Agents: Business-Savvy, Capable, Trustworthy, and Secure By 2026, the industry's competitive focus has shifted from "the quality of large language model responses" to "whether the intelligent agent can actually get things done." Haizhi's intelligent agents present a convincing case.
The most direct evidence is repurchase behavior. A large state-owned commercial bank with total assets exceeding 15 trillion yuan has collaborated with Haizhi for five years. Building on the graph database and knowledge graph platform, it has consistently repurchased intelligent agent products since 2024, with use cases now covering retail marketing, post-lending management, auditing, and operations. A major state-owned telecommunications operator has signed repurchase contracts for three consecutive years, gradually expanding from the graph database to graph analysis platforms, GraphAgent, and other ecosystem tools, forming a complete graph technology product matrix from basic storage/computation to innovative applications. Approximately 50% of the intelligent agent customers contributing to 2025 revenue were converted from existing Atlas knowledge graph clients.
Willingness to repurchase indicates that the intelligent agents are not merely conceptual or conversational but are genuinely solving business problems. So, what exactly do B2B clients need from an intelligent agent? The requirements can be summarized in four words: understanding, capability, trustworthiness, and security.
* **Business-Savvy**: The agent must have deep, built-in understanding of industry knowledge. Haizhi constructs knowledge graph ontologies as the "industry brain" for its agents, extracting tacit expert experience into structured, machine-readable contexts. For a national joint-stock bank, it built an enterprise knowledge catalog, knowledge extraction pipeline, and explicit knowledge modeling to create a unified semantic space, making knowledge AI-recognizable, inferable, and executable. * **Capable**: It must execute tasks accurately, not just offer suggestions. The company independently developed a Coding Agent technological base capable of autonomous code generation and adaptive task execution across scenarios. For an urban traffic tunnel company, it deployed agents for road network operation analysis and infrastructure maintenance, enabling network simulation, risk control, and full lifecycle infrastructure management—these are tangible actions, not just dialogue. * **Trustworthy**: The "graph-model fusion" technical approach eliminates large model hallucinations at a fundamental level, making every answer explainable and traceable. The deterministic logic of knowledge graphs complements the generalization ability of large models, ensuring reliable outputs in critical sectors like finance and energy. * **Secure**: The company pioneered the integration of a full-link security governance mechanism, deeply embedding data permission management into the graph ontology layer. This enables role-based fine-grained access control and knowledge isolation. A full-link audit system provides real-time traceability for the agent's decision paths, ensuring every action is "explainable, traceable, and auditable." The deployment of this security framework preceded industry norms, demonstrating a clear first-mover advantage.
With these four capabilities in place, the results speak for themselves: in 2025, the Atlas Intelligent Agent solution, with its 68.4% revenue growth, became the company's fastest-growing business segment. More notably are the quality metrics: the average contract value for intelligent agents reached 3.6 million yuan, higher than the 2.8 million yuan for graph solutions. The gross profit margin for intelligent agents rose to 53.2%, driving the company's overall gross margin up by 7.1 percentage points to 43.3%.
The Second Layer of Growth: Advancing into New High-Value Industries Another reason for the rapid growth of intelligent agents is the transferability of Haizhi's capabilities across industries. The company has expanded from its strongholds in finance and public security into new domains like telecommunications operators, transportation infrastructure, and unmanned systems, securing orders in these areas. Examples include building a unified graph ontology model for unmanned equipment for a management center in the Greater Bay Area and deploying an intelligent agent application platform, inclusive of large model appliances, for an overseas non-bank financial institution. The company has explicitly stated its intention to rapidly replicate its mature experience into emerging high-value fields such as advanced manufacturing, healthcare, and smart mining, while also leveraging its Hong Kong listing platform to gradually explore markets in Southeast Asia and beyond.
Why Haizhi Tech Warrants Long-Term Attention Pulling this logical chain together: robust graph computing technology supports a full-stack product matrix covering the entire digital-to-intelligent lifecycle. This comprehensive product suite fosters deep customer loyalty and high repurchase rates. The intelligent agents built upon this foundation, being business-savvy, capable, trustworthy, and secure, have been among the first in the B2B market to validate the journey from concept to repurchase. Furthermore, this capability set has the potential for cross-industry migration, opening up continuous growth avenues.
As of February 2026, the company held over 1 billion yuan in cash and financial assets, indicating ample financial reserves. The overall gross profit margin increased by 7.1 percentage points to 43.3%, and the adjusted net profit of 24.15 million yuan represented a 42.6% year-on-year increase, demonstrating sustained enhancement of profitability.
In the B2B AI landscape, the companies that truly build lasting moats are those whose solutions get the job done and earn client repurchases. Haizhi Tech's 2025 annual report delineates precisely this complete path: from technology to product, from product to competitive barrier, and from barrier to growth.
Comments