Huawei and State Grid Shaanxi Develop the Qin Xiaozhi AI Employee, Reshaping Grid Operations
Enterprise
A single voice command is all it takes. Emails are automatically sent and received, official documents drafted, visit plans generated, and outage data analyzed in seconds. This is not a glimpse into the future; it is the everyday reality for frontline staff at State Grid Shaanxi since the launch of Qin Xiaozhi, an AI-powered digital employee.
On June 24, 2026, Huawei and State Grid Shaanxi Information and Communication Company officially released Qin Xiaozhi version 1.0.0. Deployed on the utility's internal network and powered by the Huawei Atlas computing cluster, this digital co-worker is transforming daily grid operations, bringing AI capabilities directly to the intranet, where they are needed most.
State Grid Shaanxi oversees the planning, dispatching, and operation of the provincial grid, using a modern backbone network to cover the entire region. As a core entity, this network underpins regional economic growth, drives the national West-to-East Power Transmission strategy, and is central to the province's energy transition. However, the rapid expansion of wind and solar generation is making generation-grid-load-storage coordination more complex than ever, placing greater urgency on the digital transformation of grid operations.
In response, State Grid Shaanxi and Huawei forged a long-term technical innovation partnership and established a joint innovation laboratory—a benchmark for collaborative R&D between state-owned enterprises and leading technology companies. Targeting practical operational challenges, the two organizations collaborate across intelligent dispatch optimization, distribution IoT, consumption data acquisition, AI-enabled power services, and secure information infrastructure. Drawing on Huawei's portfolio of Atlas computing, AI models, GaussDB, edge computing units (ECUs), and high-speed power line communications (HPLC), they have deployed a range of innovative solutions that have demonstrated tangible benefits. These initiatives are progressively strengthening the reliability of the large-grid system, enhancing renewable integration, and accelerating Shaanxi's transition toward a new power system that is secure, balanced in supply and demand, flexible, and intelligent.
While AI is rapidly reshaping global industries, its application at the operational frontline of power grids remains fraught with practical hurdles.
1. Field staff are often required to navigate multiple, disconnected business systems throughout the day, making workflows cumbersome and inefficient.
2. A vast volume of business documentation still requires manual review and summarization, while critical institutional knowledge is frequently lost when experienced personnel rotate out.
3. Although powerful AI tools are openly available on the Internet, they cannot be directly adopted, because strict compliance requirements mandate that all power grid data remain within the internal network.
4. Compute presents an even greater bottleneck. Foundation models demand substantial computing resources, yet on-premise infrastructure is often insufficient to support large-scale model training and inference, while scaling up would require substantial capital investment and lengthy deployment cycles.
Given these challenges, to deploy AI solutions at scale, it is essential to build a high-performance computing foundation within the intranet to ensure stable operation of large models without any data leaving the secure perimeter.
(1) Atlas-Powered Joint Innovation for Breakthroughs
State Grid Shaanxi Information and Communication Company and Huawei have partnered to drive joint innovation, combining a China-developed compute foundation with deep insights into power industry use cases. Huawei contributes the Atlas computing cluster and Huawei Cloud as the technology backbone, while State Grid Shaanxi Information and Communication Company brings its accumulated domain expertise and hands-on understanding of power business operations. Together, the two parties established a joint innovation team that, after six months of concerted efforts, delivered "Qin Xiaozhi"—an AI assistant tailored for State Grid employees.
The Atlas computing cluster serves as the cornerstone of the entire solution. Powered by China-developed AI chips, it delivers robust on-premise computing capacity that fundamentally resolves the inherent tension between the mandate for data to remain within the secure intranet and the need to make AI capabilities readily accessible. Building on this Atlas-powered foundation, the two parties jointly architected a three-tier framework that covers computing foundation, intelligent engine, and scenario applications, as well as seamlessly integrating four core capabilities: large language models (LLMs), AI agents, tool invocation, and long-term memory.
(2) One AI Assistant, Multiple Roles
Leveraging Atlas's multi-model scheduling capabilities, Qin Xiaozhi can adapt to multiple roles across different job functions. For management personnel, it organizes emails and to-do lists, automatically generates task summaries, and archives important documents to a personal knowledge base, all via voice commands. For field personnel, taking the example of transformer district inspections, it connects to multiple service systems to retrieve data, identifies the target transformer districts, and generates inspection plans. A process that previously took six hours and required manual navigation across three separate systems can now be completed by Qin Xiaozhi in just ten minutes.
In addition, system prototypes can be built in one to two days using no-code development. Reports can be generated automatically based on statistical analysis of all available data. The expertise of seasoned employees can be captured and accumulated as reusable digital assets, ensuring that critical business knowledge is preserved and passed on continuously.
(3) Three Core Innovations
1. A collaborative model for skill accumulation. General office skills are centrally maintained by the project team, while specialized business skills are co-developed by frontline employees and the seed team. The innovative skills-generating-skills function creates an intranet ecosystem where anyone can become a developer.
2. The first LLM-based knowledge base within the system. Through the precise integration of three engines—knowledge framework, semantic neural network, and concept linking—Atlas compute delivers high-throughput inference to support large-scale vector retrieval. This ensures that the knowledge base responds within seconds, even after massive document ingestion. Every piece of generated content is fully traceable, eliminating AI hallucinations.
3. A self-learning and evolution mechanism. The multi-layered memory system automatically records user habits and preferences after each interaction. When encountering similar tasks, the system proactively anticipates user needs, forming a continuous learning loop of interaction, refinement, and evolution.
In the information O&M field, Qin Xiaozhi's cloud platform account audit capability has been deployed to automatically audit over 800 accounts across seven service domains, shifting account management from reactive response to proactive defense. The automated inspection coverage rate now stands at 100%, saving 1.5 hours of manual effort per day.
In system security testing, Qin Xiaozhi has reduced the time required for single-page boundary testing by 20% and improved code compliance assessment efficiency by 10%. To date, 5,940 issues have been identified and rectified through this capability.
In marketing fee calculation assurance, the team leveraged Qin Xiaozhi's database inspection skills to accurately profile the running status of microservices. This brought peak CPU usage down to around 20% and increased overall service processing efficiency by 20%.
In power distribution network repair, Qin Xiaozhi assisted in intelligent fault analysis, contributing to a 26.2% reduction in unnecessary field dispatches caused by false fault reports.
The platform now serves over 2,200 registered users, with more than 400 daily active users. Its reach spans the provincial headquarters, the Information and Communication Company, China Electric Power Research Institute, and 11 municipal subsidiaries. The platform processes over 100,000 requests per day, consuming 1.25 billion tokens. In general office scenarios, it has handled 83,000 tasks and boosted overall office efficiency by 73%. The underlying Atlas computing cluster stably supports high-concurrency inference workloads, ensuring responsive performance even during peak hours.
Qin Xiaozhi has now been widely deployed across all four levels of State Grid Shaanxi —provincial, municipal, county, and station—delivering an average productivity gain of 20%. It is estimated that the solution will unlock approximately CNY400 million in labor value province-wide and reduce system development costs by roughly CNY20 million. In high-demand scenarios, more than 80% of routine, repetitive manual tasks have been automated, while batch service processing efficiency has increased by over 60%. 100% of service response templates are now standardized. By relieving frontline staff of repetitive tasks, the solution frees them to refocus on lean governance and high-quality service delivery. Services such as electricity bill inquiries and abnormality checks are now handled within seconds, making power supply operations more efficient and consistent.
The successful deployment of Qin Xiaozhi validates the effectiveness of the joint innovation model that combines Atlas compute with deep industry expertise. Huawei provides the AI infrastructure, from chips to frameworks, while State Grid Shaanxi Information and Communication Company contributes domain knowledge and application experience. This model transforms general-purpose AI capabilities into a digital workforce that is truly usable on the frontlines of the power grid. It also offers a replicable pathway for deploying AI within intranet environments across other critical infrastructure sectors, such as transportation and finance.
From Shaanxi to the nation, and from the power sector to industries beyond, when computing power and human ingenuity work in harmony, the intranet ceases to be a security constraint and becomes a fertile ground for innovation. Technology will eventually fade into the background, unleashing the full potential of human ingenuity.