Português
EnglishA New Context Infrastructure for Agents
Recently, Qwen Office officially open-sourced a brand-new context infrastructure called “MyContext.”
The system integrates and transforms personal work data scattered across various locations and formats into specialized work files that are easier for agents to understand. This allows intelligent assistants to truly grasp user habits and actual business processes, enabling efficient human–AI collaboration in decision-making and execution, significantly enhancing performance in office scenarios.
Helping Agents Truly Understand Users and Daily Operations
Why Is Context Infrastructure Needed?
As generative AI models rapidly advance and Agent Harness architectures continue to improve, general-purpose intelligent agents have begun to deliver tangible results in boosting individual productivity. However, due to a lack of deep understanding of real-world work environments, these systems often struggle to integrate seamlessly into actual enterprise workflows, limiting overall productivity gains. According to MIT’s NANDA report, approximately 95% of enterprise-level generative AI pilots fail to generate meaningful returns, primarily because they lack sufficient data infrastructure support, preventing AI from being smoothly embedded into existing workflows.
A Dynamically Updated Digital Work File
Unlike traditional knowledge bases or vector databases, MyContext employs a localized deployment model, running directly on users’ devices. It automatically consolidates work-related information from instant messaging apps, file systems, meeting records, collaboration platforms, and multiple online and offline sources, forming a continuously evolving personalized digital file. The content covers key aspects such as job responsibilities, collaboration networks, behavioral patterns, and meeting consensus, enabling agents to quickly retrieve relevant information and perform logical reasoning when handling tasks. Each piece of data is traceable back to its original source, clearly indicating origin, time, and specific details, ensuring transparency, effectively reducing the risk of AI hallucinations, and significantly improving task accuracy and reliability.
Technical Breakthroughs Behind MyContext
In addressing challenges like time-series data and conflicting information, Qwen Office has introduced innovative algorithms and engineering designs in MyContext. For instance, when dealing with contradictory messages in instant communications, conventional approaches often rely on “taking the latest” or “choosing the most credible,” but these methods can easily result in the loss of important information. MyContext, on the other hand, first conducts semantic analysis to determine whether to merge conflicting data or prioritize one over the other; if the system cannot decide autonomously, it prompts the user for confirmation. Once approved by the user, the relevant data is locked, preventing unauthorized modifications by the model, thereby greatly safeguarding data consistency and reliability.
The Enterprise-Level Context Era Is About to Begin
Currently, MyContext supports integration with personal data from mainstream instant messaging platforms such as DingTalk and Feishu. Users simply import their data to create a dedicated smart assistant that can simulate their tone and automatically respond to work-related messages based on their personal knowledge base. Looking ahead, the development team plans to release an enterprise version of the context infrastructure, fully integrating with international enterprise systems like Salesforce and SAP while also accommodating common local workflows, ushering in an era of “enterprise context” and helping organizations achieve a leap in overall productivity.
DomTech is DingTalk’s official designated service provider in Macau, specializing in providing DingTalk services to a wide range of customers. If you’d like to learn more about DingTalk platform applications, feel free to contact our online customer service or reach out via phone at +852 95970612 or email at [email protected]. We have an excellent development and operations team with extensive market service experience, ready to provide you with professional DingTalk solutions and services!