Integrate our proprietary AI agent, MuleRun, into the core of DingTalk to enable all office functions to be automatically handled by AI, marking the official arrival of the agent-driven era in enterprise productivity software.

Article author and source: AI Frontiers Daily

Chen Yusen, the new CEO of DingTalk, issued an internal memo announcing a comprehensive overhaul of the organization, technology, and sales teams. The core focus of this adjustment is simple: integrate DingTalk’s proprietary AI agent, MuleRun, into the platform’s core infrastructure to enable all office functions to automatically operate with AI—marking the official transition of enterprise office software into an agent-driven era.

I. All Core Changes in This DingTalk Architecture Adjustment

1. The AI teams have merged, and MuleRun has become the core AI foundation of DingTalk.

Chen Yusen initially incubated the AI agent product MuleRun within Alibaba Cloud. For this initiative, the original WuKong team and the MuleRun team have been merged into a new WuKong team, with Shu Junliang, the former CTO of MuleRun, now serving as the leader. All AI-related development resources within DingTalk are now fully concentrated on iterating MuleRun, eliminating the previous分散布局 of multiple independent AI product lines.

2. The core business platform has been separately decomposed to support underlying agent execution.

The company has established a Core Platform Business Unit, led by Zhu Hong, former CTO of DingTalk, to oversee all of DingTalk’s foundational office functions. The unit’s sole core mission is to overhaul DingTalk’s underlying system so that all modules—including pages, approvals, documents, and meetings—can be directly accessed and operated by AI Agents.

3. Unified Restructuring of Sales and Back-Office Functions

Merge the six sales teams—direct sales, telesales, and third-party service providers—into a new Customer Development Department; simultaneously establish independent Marketing and Information Technology Departments. The Information Technology Department will be responsible for iterating the company’s entire business system to reduce the permissions and compatibility barriers for AI agents when accessing data and performing functional operations.

II. What office scenarios can the MuleRun Agent actually solve? Let’s break it down in plain language.

Many people confuse large models with AI agents. Here’s a simple analogy to distinguish them: A regular AI chat tool = a front desk clerk who can only execute one step at a time—you give it one instruction, and it completes just that one task. An AI agent like MuleRun = a dedicated assistant that can autonomously break down goals; once given a final objective, it plans and executes the entire workflow automatically from start to finish.

Ready-to-use office scenarios

Automatic Weekly Report Generation and Synchronization: Employees upload chat logs from work groups, meeting minutes, and project documents. MuleRun automatically extracts relevant work content, consolidates project data, and generates standardized weekly reports, which can be sent with one click to the corresponding direct managers—eliminating the need for manual compilation and copying of materials.

Cross-departmental approval self-tracking: After employees initiate processes such as procurement, leave requests, or project proposals, the Agent monitors approval nodes in real time, automatically sending reminders to overdue approvers and consolidating all feedback into the original document—eliminating the need for manual follow-ups online or offline.

End-to-end autonomous meeting management: Before the meeting, the Agent gathers relevant project materials to generate an agenda; during the meeting, it transcribes speech in real time and identifies key points from each speaker; after the meeting, it automatically breaks down action items, assigns them to responsible parties, and sets deadlines.

Customer documents are automatically archived and organized. Chat records, collaboration contracts, and call recordings generated by sales and customers are automatically tagged and categorized by the Agent, which extracts key information such as customer requirements and collaboration quotes for storage. Subsequent colleagues can retrieve the entire communication history with a single click.

MuleRun's core advantages over general AI tools

Natively integrated with DingTalk's underlying database, eliminating the need for users to manually upload files—directly access internal data such as corporate contact lists, approval workflows, cloud documents, and meeting recordings.

Supports multi-step, continuous, autonomous execution of complex tasks without requiring users to issue multiple separate commands.

The supporting IT department has continuously optimized system interfaces, significantly reducing permission errors and format incompatibility issues when AI calls office functions.

III. Industry Status Breakdown: The Competitive Logic of the Office Sector Has Completely Transformed

1. The office AI sector has completed its phase transition.

In previous years, enterprise service providers competed on digital capabilities, primarily migrating offline paper forms and manual approval processes to online systems. Today, the core of industry competition has shifted to native intelligent capabilities, with vendors vying to deliver products that support AI-driven autonomous operating functions and reduce the frequency of manual, repetitive tasks. DingTalk’s recent structural adjustment directly anchors enterprise collaboration software competition in the foundational implementation of AI agents.

2. Two distinctly different technological approaches by major office equipment manufacturers

DingTalk Roadmap: Underlying Reconstruction + Proprietary Agent Base Built on MuleRun to Establish a Unified AI Capability Foundation, Rebuilding the Office System from the Ground Up, with All Existing Office Modules Redesigned Around Agents, Making AI the Core Carrier of Product Operation.

Most collaborative tools follow this approach: they add AI features on the surface while keeping the existing mature underlying architecture unchanged, merely introducing an AI chat interface on the software page and enabling only lightweight functions such as document summarization and text-based Q&A, with AI serving as an optional value-added module.

The two approaches create a significant difference in usability: external AI can only handle single text-based tasks, while native agents can autonomously link multiple business modules to complete an entire workflow.

3. Three Perceptible Changes in Future Enterprise Office Tools

Regular employees don’t need to frequently switch between software functions—just set the final work objective, and AI will automatically connect modules such as documents, approvals, meetings, and customer management to complete the task.

When enterprises procure collaborative software, they no longer focus solely on basic features such as attendance, forms, and video conferencing; instead, they prioritize verifying whether the platform’s underlying infrastructure can reliably support AI agents.

The responsibilities of internal IT roles have shifted, requiring not only ensuring system stability but also continuously optimizing various business interfaces to accommodate data retrieval and function invocation needs of AI agents.

Four, Future Ideas

This DingTalk organizational restructuring does not include any short-term commercialization plans; all human and technical resources are being directed toward the foundational overhaul of MuleRun. All future feature updates for DingTalk will be developed by the newly integrated AI team. The foundational transformation of office software AI agents will become the industry’s unified direction, and other competitors will progressively follow suit with upgrades to their underlying architectures to support large-scale Agent deployment.