DingTalk and Feishu have evolved far beyond messaging apps. In 2026, they function as the central nervous system for millions of Chinese enterprises — and both platforms are building increasingly sophisticated AI ecosystems. For organisations looking to deploy conversational BI, understanding how to integrate with these platforms is no longer optional; it is the primary deployment model in China.
Key Insight: DingTalk and Feishu AI ecosystems enable enterprises to deploy conversational BI directly inside the IM tools that employees already use daily, eliminating adoption barriers and delivering data insights through natural language queries in the workflow.
The Two Titans of Chinese Enterprise IM
China's enterprise collaboration market is dominated by two platforms, each backed by a technology giant:
- DingTalk (Alibaba Group): With over 700 million individual users and 25 million enterprise organisations, DingTalk is the most widely deployed enterprise IM platform in China. It has particularly strong penetration in manufacturing, retail, and government sectors.
- Feishu / Lark (ByteDance): Feishu (known internationally as Lark) has rapidly gained market share, particularly among technology companies, internet firms, and knowledge-intensive industries. Its document-first design philosophy and strong third-party integration ecosystem make it popular with startups and scale-ups.
DingTalk's AI Assistant Ecosystem
DingTalk has invested heavily in its AI capabilities through several key initiatives:
- DingTalk AI Assistant Platform: Launched in 2023 and significantly expanded through 2026, this platform allows enterprises to build custom AI assistants that operate within DingTalk chat threads. It supports integration with Alibaba's Tongyi Qianwen (Qwen) LLM as well as third-party models.
- Enterprise Agent Market: A marketplace where enterprises can discover, deploy, and share pre-built AI agents. As of mid-2026, the marketplace hosts thousands of agents covering HR, finance, operations, and analytics functions.
- Action-based architecture: DingTalk's AI assistant framework supports "actions" — discrete API calls that the AI can trigger based on user requests. This is architecturally similar to MCP's tool-calling paradigm, making DingTalk a natural fit for MCP-powered conversational BI.
Feishu's AI Capabilities
Feishu has taken a different but equally powerful approach:
- Feishu Intelligent Partner (Feishu Bot 2.0): Feishu's AI bot framework supports multi-turn conversations, context awareness across chat threads, and the ability to invoke external APIs. It supports ByteDance's Doubao model as well as other domestic and international LLMs.
- Feishu AnyCross: An integration platform that connects Feishu to over 300 enterprise applications. AnyCross provides pre-built connectors for ERP, CRM, and database systems — complementing MCP's data source connectivity.
- Document AI: Feishu's strength in collaborative documents extends to AI-powered document analysis, automated report generation, and data extraction from spreadsheets — capabilities that pair naturally with conversational BI for ad-hoc analysis.
MCP Compatibility with Chinese IM Platforms
The Model Context Protocol is fundamentally a client-server protocol that defines how AI models access data. Integrating MCP with DingTalk or Feishu requires a bridge layer — an adapter that translates between the IM platform's bot framework and the MCP server. Here is how it works:
- IM Bot receives a user query in a DingTalk group chat or Feishu thread.
- The bridge adapter (developed by Beehive Strategy or your integration team) receives the message via the IM platform's webhook or API.
- The LLM processes the query and determines which MCP servers need to be called to retrieve relevant data.
- MCP servers execute against enterprise data sources (MySQL, Snowflake, SAP, etc.) and return structured results.
- The LLM formats the response — including charts, tables, or natural language summaries — and the bridge adapter posts it back to the IM conversation.
Both DingTalk and Feishu support rich message formats (cards, interactive elements, chart embeds), which means conversational BI responses can include visually formatted data visualisations, not just plain text.
Platform Comparison for BI Deployment
| Dimension | DingTalk | Feishu / Lark |
|---|---|---|
| Enterprise User Base | 25M+ organisations | 12M+ organisations |
| Primary Sectors | Manufacturing, retail, government | Technology, internet, professional services |
| Default LLM | Qwen (Tongyi Qianwen) | Doubao (ByteDance) |
| Third-Party LLM Support | Yes (via API) | Yes (via API) |
| Bot Framework | AI Assistant Platform | Feishu Bot 2.0 |
| Rich Message Cards | Yes | Yes |
| Enterprise App Market | DingTalk Open Platform | Feishu App Store |
| MCP Integration | Via bridge adapter | Via bridge adapter |
Practical Implementation Steps
For enterprises looking to deploy conversational BI through DingTalk or Feishu, the implementation path is straightforward:
- Choose your IM platform. If your organisation already standardises on one, start there. If you operate across both, a unified backend with platform-specific adapters is the recommended architecture.
- Register an enterprise bot. Create a bot application through the DingTalk Open Platform or Feishu Developer Console. Configure webhook endpoints and permission scopes.
- Deploy the MCP bridge. Deploy the bridge adapter that connects your IM bot to the MCP server infrastructure. This handles message routing, authentication, and response formatting.
- Configure data connectors. Set up MCP servers for your key enterprise data sources — ERP, CRM, financial systems, and data warehouses.
- Test and iterate. Start with a pilot group of 10-20 power users. Collect feedback on query accuracy, response speed, and user experience before scaling organisation-wide.
The Bottom Line
DingTalk and Feishu are not just messaging platforms — they are the operating system for enterprise work in China. Deploying conversational BI through these platforms means meeting your users where they already work, eliminating the adoption friction that kills standalone analytics tools.
At Beehive Strategy, we provide pre-built integrations for both DingTalk and Feishu, enabling you to deploy MCP-powered conversational BI directly into your team's daily workflow within days, not months. Book a free demo to see how.
Frequently Asked Questions
How does conversational BI integrate with DingTalk and Feishu?
Conversational BI integrates with DingTalk and Feishu through their bot APIs. The AI agent connects to the enterprise semantic layer and responds to natural language queries within group chats or direct messages, delivering charts, tables, and narrative insights without requiring users to switch to a separate BI tool.
What are the benefits of IM-native BI over standalone BI tools?
IM-native BI eliminates the adoption barrier of switching to a separate application. Since employees already spend significant time in DingTalk or Feishu, delivering insights directly in their workflow dramatically increases engagement. Organisations deploying IM-native conversational BI see 3-5x higher adoption rates compared to standalone BI dashboard tools.
How do you ensure data security when using IM-based BI?
Data security is enforced at multiple levels: the semantic layer controls which data each user role can access, all queries are logged for audit, data never leaves the enterprise's controlled environment (the IM bot only receives pre-formatted results, not raw data), and optional row-level security ensures users only see data they are authorised to access.