Feishu New Product Launch Deep Analysis: The Leap Towards an Agent-Native Platform for Chinese Enterprises

The core question this section aims to answer: What revolutionary changes did the Feishu new product launch bring, and why does this represent a significant leap for Chinese enterprise software towards an Agent platform?

Feishu has just completed a highly significant new product launch. As a technology observer with long-term focus on enterprise-level AI applications, I followed the entire event and conducted an in-depth experience analysis. The most fundamental change in this launch is the comprehensive iteration of the existing AI assistant Aily and Feishu Miaoda, moving towards a truly Agent-native platform.

If you have experienced Claude’s product line, you can quickly understand the positioning of this upgrade: Feishu Aily is more like Cursor, or can be understood as an enterprise version of OpenClaw running directly within Feishu; while Feishu Miaoda is similar to Claude Code. One is an assistant Agent that solves daily work problems, the other is a professional development Agent.

With this launch, Feishu has truly become the most convenient platform in China for using Agents. Let me analyze these three core upgrades in detail.

1. Feishu Aily: Evolution from AI Assistant to Enterprise-Level Agent

The core question this section aims to answer: What are the fundamental differences between Feishu Aily and its predecessor? How does it address the pain points enterprise users face in actual work scenarios?

1.1 Core Upgrade: From Q&A Tool to Resident Agent

The most fundamental change in Feishu Aily’s upgrade is its transformation from a “you ask, it answers” AI assistant into an intelligent Agent that resides in the Feishu contact list. In terms that are easiest to understand today, it has become Feishu’s own completely built-in, ready-to-use OpenClaw.

Core Advantages:

  • Native support for accessing all enterprise data without additional configuration
  • Inherits Feishu’s hierarchical user permission system
  • Direct access to your messages, documents, multi-dimensional tables, schedules, and tasks

Activation and Configuration Process:

  1. Open the left sidebar in Feishu and find “Smart Partner”
  2. Click “Activate Now” and wait a few dozen seconds to start the conversation
  3. The system automatically recognizes user identity, department, and group information
  4. Further personalize settings, similar to configuring personality for Claude

From a technical implementation perspective, this native integration avoids the cumbersome deployment processes of traditional solutions: no longer needing to separately deploy Claude bots, no need to connect Feishu plugins, no need to handle complex authorization issues. The Agent directly “lives” in Feishu and has natural access to all user work data.

1.2 Practical Application Scenarios: From Weekly Reports to Community Management

Scenario 1: Intelligent Weekly Work Report Generation

Previously, we needed to manually organize a week’s work content: reviewing chat records, meeting minutes, task lists, and then compiling them into structured reports. Now, simply tell Aily: “Help me see what I’ve done this week,” and it automatically:

  • Reads work discussions from chat records
  • Analyzes key decisions from meeting minutes
  • Summarizes task list completion status
  • Organizes into clearly formatted weekly report documents

Technical Implementation Details:

# Aily internal workflow示意
def generate_weekly_report(user_id):
    # 1. Get data within user's permission scope
    messages = get_chat_messages(user_id, time_range="week")
    meetings = get_meeting_minutes(user_id, time_range="week")
    tasks = get_task_list(user_id, status="completed")
    
    # 2. Intelligent analysis and integration
    work_summary = analyze_work_patterns(messages, meetings, tasks)
    
    # 3. Generate structured report
    report = format_structured_report(work_summary)
    return report

Scenario 2: Community Member Status Management Automation

Our community operations team member previously managed community applications manually: after sending invitations, they needed to search for WeChat names in group member lists, confirm who joined the group, update statuses, and then delete notes. This consumed at least one hour daily.

Aily’s solution:

  1. Send multi-dimensional tables and two community member lists to Aily
  2. Tell it the fields that need matching and update logic
  3. Complete matching and updating of 609 data entries within seconds
  4. Automatically statistics processing results: successful matches, unmatched entries

Data Comparison Table:

Task Type Traditional Method Time Aily Processing Time Efficiency Improvement Factor
Community Member Status Update 60+ minutes 2 minutes 30x
Weekly Report Organization 30-45 minutes 30 seconds 60-90x
Data Matching Verification Depends on data volume Real-time processing Non-quantifiable

1.3 Skills Ecosystem: Expanding Capability Boundaries

Aily supports installing various Skills to expand capability boundaries. The official SkillHub platform includes Skills with security scanning and open-source certification. Users can install different skills for Aily as needed:

Common Skill Categories:

  1. Data Operation Skills: Multi-dimensional table operations, database queries
  2. Content Processing Skills: Document generation, web scraping, public account analysis
  3. Process Automation Skills: Scheduled tasks, approval workflows, data synchronization

Skills Migration Process:

  1. Export Skills list from old Claude bot
  2. Forward to Aily for installation
  3. Installation process takes approximately 10 minutes

Author’s Reflection: The construction of the Skills ecosystem makes me think about a crucial point – the success of enterprise-level AI applications depends not only on core model capabilities but even more on ecosystem maturity. Feishu’s native Skills system allows users to expand Agent capabilities like installing mobile apps. This low-barrier expansion approach is key to large-scale enterprise adoption.

2. Feishu Miaoda: The Qualitative Leap from Low-Code to Code Agent

The core question this section aims to answer: How has Feishu Miaoda evolved from a traditional low-code platform to a true Code Agent? What practical impact does this have on enterprise development workflows?

2.1 Underlying Architecture Upgrade: Native Code Agent

The most exciting upgrade for me in Miaoda is the transformation of its underlying architecture into a native Code Agent. Previously, when developing bots on the Feishu developer platform using Claude Code, the process was quite painful: various permission issues, authorization problems, and API call challenges constantly emerged.

Because Claude Code didn’t understand the Feishu ecosystem, I needed to manually provide Feishu API documentation each time, explaining user permissions, application permissions, and other concepts. Now, Miaoda natively supports the Feishu ecosystem, providing a qualitative leap in development experience.

Development Entry Point: In the Aily work assistant sidebar, click “Develop Application” to enter the Miaoda interface. Two modes are available:

  • Inspiration Exploration Mode: Suitable for creative brainstorming and prototype design
  • Application Development Mode: For developing actually usable applications

2.2 Practical Case: Automated Contract Filling System

Our business department colleagues have long struggled with contract filling work: after legal prepares contract templates, business colleagues need to create copies, then find 29 field positions one by one for filling. This type of work has low technical complexity but consumes enormous energy and is prone to errors.

Miaoda Development Process:

  1. Requirements Analysis Phase:

    • Send requirement descriptions to Miaoda
    • Miaoda first creates a preliminary design, confirming understanding is correct
    • Generates detailed technical plan (though users might not read it carefully)

  2. Development Execution Phase:

    // Core code logic示意 generated by Miaoda
class ContractFiller {
  constructor(templatePath, fieldMapping) {
    this.template = loadTemplate(templatePath);
    this.mapping = fieldMapping;
  }
  
  fillContract(userData) {
    // 1. Create contract copy
    const contract = this.template.createCopy();
    
    // 2. Intelligent field matching and filling
    Object.keys(this.mapping).forEach(field => {
      const value = userData[this.mapping[field]];
      contract.fillField(field, value);
    });
    
    // 3. Generate preview and export options
    return {
      preview: contract.generatePreview(),
      download: contract.exportToPDF(),
      share: contract.shareToUsers(userData.recipients)
    };
  }
}

  3. Final Deliverables:

    • Complete contract filling interface
    • Real-time preview functionality after inputting information
    • One-click export and download functionality
    • Support for selecting recipients and sending directly to Feishu

Technical Implementation Highlights:

  • Intelligent Field Matching: Even with inconsistent field names, semantic understanding enables matching
  • Permission Auto-inheritance: When sharing applications, permission systems automatically expand
  • Real-time Collaboration: Supports multiple developers collaborating on the same application

2.3 Deep Integration with Aily: End-to-End Automation

Further optimization involves integrating with Aily: placing master contract fields in cloud documents, allowing Aily to automatically read and fill them. This eliminates even manual field entry, achieving true end-to-end automation.

Complete Workflow:

  1. Business personnel maintain partner basic information in cloud documents
  2. Aily periodically reads updates and synchronizes with the contract system
  3. When contracts are needed, the Miaoda-generated system automatically fills all fields
  4. Legal reviews online, and after approval, automatically sends to partners

Author’s Reflection: This deep integration makes me profoundly realize that individual AI tools have limited capabilities, but the synergistic effect between tools can create exponential value. By building an ecosystem of Aily (work assistant) and Miaoda (development platform), Feishu enables enterprise users to complete the full cycle from requirement proposal to application development to daily use on a single platform.

3. Multi-dimensional Table Agent: The Intelligent Revolution in Data Work

The core question this section aims to answer: How does the Multi-dimensional Table Agent solve data analysis and visualization pain points? What practical value does it offer non-technical users?

3.1 From Data Storage to Intelligent Analysis

Multi-dimensional tables are tools that every Feishu user inevitably encounters. While highly efficient, truly mastering them has significant learning curves. For example, I previously created a small tool that daily scrapes public account data into multi-dimensional tables.

Traditional Pain Points:

  • Dense data makes trends difficult to visualize intuitively
  • Manual creation of various charts and dashboards required
  • Time-consuming analysis processes prone to errors

Agent Solution:

  1. Create table copies (protecting original data)
  2. Simply tell the Agent: “Help me create a data dashboard”
  3. Agent confirms requirements, decomposes tasks, executes analysis
  4. Generates complete visual dashboard within minutes

3.2 Practical Application: Public Account Data Analysis Dashboard

My personal public account data scraping tool generates nearly 500 detailed data entries daily, including metrics like read counts, likes, shares, and user sources across multiple dimensions.

Agent-Generated Analysis Content:

  1. Trend Analysis Charts:

    • Read count time series line charts
    • Different article type performance comparison bar charts
    • User active time period heatmaps

  2. Intelligent Insights:

    • Identification of best-performing article types
    • Analysis of user preference change trends
    • Future content direction suggestions

  3. Reusable Templates:

    • All charts and analyses can be copied and transferred
    • Support for one-click application to other data tables
    • Customizable modification options

Technical Architecture:

class MultiTableAgent:
    def __init__(self, table_data):
        self.data = table_data
        self.analysis_tools = {
            'trend': TrendAnalyzer(),
            'comparison': ComparisonEngine(),
            'visualization': ChartGenerator()
        }
    
    def create_dashboard(self, requirements):
        # 1. Requirement understanding and task decomposition
        tasks = self.decompose_requirements(requirements)
        
        # 2. Parallel execution of analysis tasks
        results = self.execute_parallel_tasks(tasks)
        
        # 3. Generate visual dashboard
        dashboard = self.compile_dashboard(results)
        
        # 4. Add intelligent analysis suggestions
        insights = self.generate_insights(dashboard)
        
        return {
            'dashboard': dashboard,
            'insights': insights,
            'reusable_components': self.extract_reusable_components()
        }

3.3 Value for Non-Technical Users

Lowering Technical Barriers:

  • No need to learn SQL or data analysis tools
  • Natural language description of requirements yields professional analysis
  • Automatically generates readable reports and charts

Improving Decision Quality:

  • Data-based objective analysis reduces subjective judgment
  • Real-time updated data dashboards identify issues promptly
  • Intelligent alerts and trend prediction functionality

Author’s Reflection: The significance of the Multi-dimensional Table Agent extends beyond technical implementation to democratizing data analysis. In traditional enterprises, data analysis is typically the domain of specialized teams, with ordinary business personnel rarely participating. Now, any frontline employee can converse with data through natural language. This change will profoundly impact organizational decision-making processes.

4. The New Paradigm of Enterprise Software in the Agent Era

The core question this section aims to answer: Why should the entry point for enterprise software in the Agent era be the daily work platform? What insights does Feishu’s leap provide for the future?

4.1 Three Essential Elements for Agent Implementation Success

I’ve always believed that the true entry point for the Agent era isn’t app stores or chatbots, particularly for efficiency-focused scenarios. It’s most likely the platform you use daily for work. For Agents to truly help you work, they need three things:

First, Context
Agents need to know who you are, what you’re doing, who your colleagues are, what projects you’re currently pushing, what your boss cares about. Without this context, Agents are essentially guessing. Feishu naturally possesses complete work context that external AI tools struggle to obtain.

Second, Tools
Agents need to actually operate things – not just provide suggestions, but truly edit documents, send messages, create tables, schedule meetings, submit approvals. Agents that only talk without action have no practical value. Feishu provides a complete tool ecosystem that Agents can seamlessly call.

Third, Permissions
In enterprise scenarios, who can see what data, which operations require approval – these must be strictly controlled. An Agent without a permission system is unusable in enterprises. Feishu’s mature permission system can be directly inherited by Agents.

4.2 Feishu’s Competitive Advantage: Ecosystem Closure

Feishu’s greatest advantage lies in constructing a complete enterprise work ecosystem closure:

  1. Communication Layer: Instant messaging, group chats, meetings
  2. Collaboration Layer: Documents, tables, calendars, tasks
  3. Development Layer: Miaoda platform, open APIs
  4. AI Layer: Aily Agent, Multi-dimensional Table Agent

This closed ecosystem enables Agents to:

  • Understand requirements through communication
  • Execute tasks through collaboration
  • Expand capabilities through development
  • Continuously learn and optimize at the AI layer

4.3 Insights for Future Enterprise Software Development

Insight 1: Native Integration Outperforms External Connections
This Feishu upgrade proves that AI capabilities natively integrated into work platforms provide far superior experiences compared to external tool connections. This reduces configuration complexity, lowers usage barriers, and enhances security.

Insight 2: Permission Systems Are the Foundation of Enterprise AI
Without comprehensive permission management, enterprise-level AI applications cannot be implemented. Feishu directly empowers Agents with its mature permission system, addressing enterprises’ most critical security concerns.

Insight 3: Ecosystem Construction Outweighs Point Breakthroughs
Individual AI functions, no matter how powerful, struggle to generate actual business value without complete ecosystem support. By building a synergistic ecosystem of Aily, Miaoda, and Multi-dimensional Table Agents, Feishu creates a 1+1+1>3 effect.

Author’s Reflection: Observing Feishu’s upgrade brings to mind an analogy: failing to raise lobsters isn’t because the lobsters are inadequate, but because the pond lacks water. Many enterprises attempting AI transformation focus solely on AI models themselves while neglecting the infrastructure supporting AI effectiveness. Through years of accumulation, Feishu has constructed a “water-filled pond” – this is the prerequisite for AI to truly deliver value.

5. Implementation Guide and Best Practices

The core question this section aims to answer: How can enterprises effectively implement Feishu’s Agent capabilities? What are the specific implementation steps and considerations?

5.1 Phased Implementation Roadmap

Phase 1: Personal Assistant Trial (1-2 weeks)

  • Activate Aily Smart Partner for team members
  • Experience basic functions: weekly report generation, schedule management, information queries
  • Collect usage feedback and identify pain points

Phase 2: Team Collaboration Expansion (2-4 weeks)

  • Establish department-level Skills libraries
  • Develop team-specific tools and workflows
  • Train team members to create simple applications using Miaoda

Phase 3: Business Process Reconstruction (1-2 months)

  • Identify high-value automation scenarios
  • Use Miaoda to develop core business applications
  • Establish AI-assisted decision support systems

Phase 4: Organizational Culture Transformation (Ongoing)

  • Establish AI capability certification systems
  • Cultivate internal AI development experts
  • Build AI-driven innovation culture

5.2 Key Success Factors

  1. Leadership Support: Senior management must understand the strategic significance of AI transformation
  2. Pilot Project Selection: Begin with high-value, low-risk scenarios
  3. User Training System: Establish tiered and categorized training programs
  4. Effect Evaluation Mechanism: Quantify efficiency improvements from AI applications
  5. Continuous Optimization Cycle: Continuously improve applications based on usage feedback

5.3 Common Issues and Solutions

Issue 1: Data Security and Privacy Protection

  • Solution: Leverage Feishu’s existing permission system; Agents only access data within user permission scope
  • Implementation Suggestion: Start with low-sensitivity data, gradually expand as trust builds

Issue 2: User Acceptance and Habit Formation

  • Solution: Provide comprehensive training and support, demonstrate actual value cases
  • Implementation Suggestion: Establish “AI Champion” roles, with early adopters assisting colleagues

Issue 3: Technical Barriers and Maintenance Costs

  • Solution: Utilize Miaoda’s low-code characteristics to reduce development barriers
  • Implementation Suggestion: Establish internal Skills sharing libraries to reduce duplicate development

6. Future Outlook and Industry Impact

The core question this section aims to answer: What impact does Feishu’s upgrade have on the entire enterprise software industry? What are the future development trends?

6.1 Industry Competition Landscape Changes

Feishu’s leap will redefine the competitive dimensions of enterprise software:

Traditional Competition Dimensions:

  • Functional completeness
  • User experience
  • Pricing strategy
  • Customer service

New Competition Dimensions:

  • AI-native degree
  • Agent ecosystem
  • Intelligent automation capability
  • Business understanding depth

6.2 Technology Development Trend Predictions

Short-term Trends (1-2 years):

  • More enterprise software platforms will integrate Agent capabilities
  • Industry-specific Agent solutions will emerge
  • Agent collaboration protocols will become standardized

Medium-term Trends (3-5 years):

  • Agents become standard configuration in enterprise software
  • New workflow types optimized specifically for Agents will emerge
  • Agents will begin participating in strategic decision support

Long-term Trends (5+ years):

  • Human-Agent collaboration becomes the mainstream work mode
  • Enterprise organizational structures will be reconstructed due to Agents
  • New business models and competitive strategies will emerge

6.3 Implications for Enterprises and Individuals

Implications for Enterprises:

  1. Strategic Level: Incorporate AI capability building into core strategy
  2. Organizational Level: Cultivate AI literacy, adjust organizational structure
  3. Technical Level: Choose platforms with AI-native capabilities
  4. Cultural Level: Establish innovation-embracing corporate culture

Implications for Individuals:

  1. Skill Development: Learn new skills for collaborating with AI
  2. Work Approach: Delegate repetitive work to Agents
  3. Career Planning: Transition toward creative work that AI cannot easily replace
  4. Thinking Mode: Cultivate systematic thinking and problem-solving abilities

Author’s Reflection: Reviewing Feishu’s upgrade, I see not just a product iteration but a microcosm of the entire industry advancing toward the AI-native era. In this transformation, the biggest winners may not be the enterprises that first launch AI functions, but those that best understand business essence, can construct complete ecosystems, and minimize usage barriers. Through this leap, Feishu demonstrates the correct evolutionary path for an enterprise software platform in the AI era.

Practical Summary and Action Checklist

Core Value Summary

  1. Aily Smart Partner: Evolved from Q&A tool to resident work assistant, natively integrated with enterprise data
  2. Miaoda Code Agent: Upgraded from low-code platform to true development Agent, lowering application development barriers
  3. Multi-dimensional Table Agent: Democratizes data analysis, enabling non-technical users to gain professional insights
  4. Complete Ecosystem: Constructs a closed loop of communication, collaboration, development, and AI

Quick Implementation Checklist

✅ Week 1: Activate Aily, experience basic functions, collect team requirements
✅ Week 2: Install common Skills, establish department Skills libraries
✅ Week 3: Develop first practical application using Miaoda, address specific pain points
✅ Week 4: Establish effect evaluation system, optimize usage experience
✅ Continuous Optimization: Iteratively improve based on feedback, expand application scenarios

One-page Summary

Feishu Agent Platform Core Capabilities:

  • Aily: Intelligent work assistant, supports enterprise data access, task automation, Skills expansion
  • Miaoda: Code Agent development platform, low-code application development, deep Feishu ecosystem integration
  • Multi-dimensional Table Agent: Intelligent data analysis, visual dashboards, trend prediction

Implementation Key Success Factors:

  • Leadership support and strategic prioritization
  • Begin with high-value scenario pilots
  • Establish tiered training systems
  • Quantify effects for continuous optimization

Expected Benefits:

  • Repetitive work efficiency improvement: 30-60x
  • Development barrier reduction, business personnel can independently develop tools
  • Data analysis democratization, improving decision quality
  • Building AI-driven innovation culture

Frequently Asked Questions (FAQ)

Q1: What’s the difference between Feishu Aily and OpenClaw?
A: Aily is Feishu’s native intelligent assistant, directly integrated into the enterprise work platform, requiring no additional configuration to access all enterprise data. OpenClaw is an external tool requiring complex deployment and permission configuration. Aily has clear advantages in security, usability, and integration level.

Q2: Can non-technical personnel use Miaoda to develop applications?
A: Absolutely. Miaoda uses natural language interaction; users simply describe requirements, and the Agent automatically completes requirements analysis, technical design, and code implementation. Even without programming experience, practical work applications can be developed.

Q3: Will Agents see my private chat records?
A: No. Agents follow Feishu’s existing permission system and can only access data you have permission to access. Private chats, confidential documents, and other permission-protected content remain inaccessible to Agents.

Q4: Do these AI functions require additional payment?
A: Basic functions are typically included in Feishu standard services, while advanced functions and enterprise-level features may require corresponding authorization. Specific fee structures should be consulted with Feishu official sales representatives.

Q5: How is data security ensured during Agent processing?
A: Feishu employs multi-layer security mechanisms: data transmission encryption, permission access control, operation audit logs, compliance certification, etc. All Agent operations occur within Feishu’s security framework.

Q6: How should enterprises begin using these functions?
A: We recommend phased implementation: 1) Activate Aily to experience basic functions; 2) Train key users; 3) Identify high-value scenario pilots; 4) Gradually expand organization-wide. Feishu official provides complete implementation guidance and support.

Q7: Where do Skills come from? How is security ensured?
A: Skills primarily come from the official SkillHub platform, with all Skills undergoing security scanning and open-source certification. Enterprises can also develop internal private Skills to ensure compliance with their security and regulatory requirements.

Q8: Can Agents replace human work?
A: Agents aim not to replace humans but to enhance human capabilities. They handle repetitive, rule-based work, allowing humans to focus on creative, strategic tasks. In practical applications, human-Agent collaboration efficiency far exceeds working alone.

Through this deep analysis, we can see Feishu leading Chinese enterprise software in a significant leap toward Agent-native platforms. This isn’t merely a technical upgrade but a transformation of work methods and enterprise operation models. For enterprises already using Feishu, now is the time to actively explore these new capabilities; for enterprises not yet using Feishu, this may present an important opportunity to reconsider digital strategy.

The true AI era isn’t the future – it’s happening now. And the wisest approach is to choose an already “water-filled pond,” allowing AI seeds to take root, sprout, and thrive here.

Image Source: Unsplash (enterprise collaboration scenario示意 images)