Not office assistants — three different philosophies of enterprise digitalization.
In 2026, China’s tech giants are running their enterprise-agent race white-hot. Tencent shipped WorkBuddy; Alibaba embedded Tongyi Qwen deep into DingTalk and its cloud; ByteDance built the complete Trace Work agent platform on top of Feishu (Lark). The surface contest looks like “whose AI assistant is nicer to use,” but study the products and the truth emerges: they represent three entirely different philosophies of enterprise digitalization.
This is not three runners in one race. It is three different races that happen to share a name. Understanding the differences means understanding each company’s core competence — none of them is “doing AI office tools”; each is using its core strength to redefine what office work is.
Tencent WorkBuddy: Rebuilding Business Flow on the Relationship Graph
WorkBuddy holds one asset no competitor can replicate: the relationship graph between WeChat and WeCom (Enterprise WeChat) covering 1.3 billion people. This is not an office product — it is an agent platform that rewrites business processes in social relationships.
Once a WorkBuddy agent has a customer’s WeChat ID, it can do far more than message them: it sees the customer’s chat history with the brand, past interactions, purchase preferences; it reaches customers across WeCom and personal WeChat; it runs automated operations inside WeChat groups — all on Tencent’s unique social graph. Three layers of capability:
First, relationship-graph reach. The WeCom-to-WeChat channel into 1.3 billion users is a gold mine on its own. WorkBuddy agents run true end-to-end customer lifecycle management — acquisition, operations, repurchase — closed entirely inside the WeChat ecosystem. Competitors can copy the agent features; they cannot copy the graph.
Second, cross-domain orchestration of official accounts, mini programs, and WeCom. Tencent has connected official-account content operations, mini-program transaction loops, and WeCom customer management. WorkBuddy agents flow between these touchpoints automatically — spotting high-intent behavior from an official-account follower, then adding them on WeCom and pushing a personalized mini program. Content → social → transaction, seamless.
Third, Tencent Cloud plus Hunyuan underneath. The Hunyuan foundation model plus Tencent Cloud’s enterprise services pushes WorkBuddy toward a “large-enterprise custom delivery” model: we give you the model, the platform, the ecosystem — you configure agents inside your own business flows. The positioning is clear: WorkBuddy does not help enterprises work more efficiently; it helps them redesign how they reach customers. The competitive edge is not the model — it is the largest social network in China standing behind it.
Alibaba Tongyi: Infrastructure-Driven Productivity
If Tencent’s edge is social relations, Alibaba’s is infrastructure depth. Tongyi Qwen for Work takes a categorically different position: it does not start from “reaching customers” but from “internal organizational productivity” — positioned as the AI-native replacement for traditional office suites.
The competitive structure behind it: cloud infrastructure is the base. Alibaba Cloud carries Tongyi’s training and inference. For an enterprise agent handling heavy documents, spreadsheets, and collaboration, cloud elasticity, cost, and latency directly determine the experience — Alibaba Cloud’s 100+ availability zones, its CDN, and the Tongyi models are deeply bundled, and every AI feature inside DingTalk is Alibaba Cloud shipping compute. DingTalk is the enterprise entry point. Its collaboration-market share in China is roughly level with Feishu, and Tongyi agents deploy directly into DingTalk workflows: approvals, attendance, document collaboration, meetings — the highest-frequency enterprise behaviors are Alibaba’s most natural landing zones. E-commerce and local-life data close the loop. The overlooked asset: Taobao, Tmall, and Amap’s massive transaction data. Tongyi for Work is not just an office assistant — it can be a brand’s intelligent operations analyst, parsing sales data, optimizing supply chains, generating marketing strategy. That is breadth WorkBuddy lacks and depth Trace Work cannot reach.
Alibaba’s strategy in one line: use cloud scale to cut agent costs, DingTalk penetration to own the enterprise entrance, and e-commerce data to differentiate. Its agent is not “a smarter assistant” — it is “an AI coprocessor embedded in the organization’s operating system.”
ByteDance Trace Work: Knowledge-Flow Collaboration
ByteDance’s Trace Work walks a completely different road — not optimizing existing office processes but redefining how knowledge moves. Feishu is a document-centric collaboration platform: unlike DingTalk’s approvals-and-controls or WeCom’s customer-reach, Feishu’s DNA is information flow and knowledge management — ByteDance runs tens of terabytes of documents and knowledge bases on it internally, and Trace Work grew out of that soil.
Three dimensions set it apart. First, documents as agents. In Trace Work’s frame, every Feishu document can be “understood” and converted into executable action: publish a PRD and Trace Work parses the task assignments, deadlines, and stakeholders inside it, creating task flows, reminders, and notifications automatically. Documents stop being passive containers and become active business drivers. Second, knowledge-graph-driven agent collaboration. ByteDance’s core recommendation-algorithm capability — structured understanding and personalized distribution of massive information — migrates into enterprise knowledge management: Trace Work agents build dynamic knowledge graphs from Feishu docs, chats, and meeting notes, then push the most relevant information and action suggestions per employee, per role, per context. Third, zero-code agent building. ByteDance’s low-code platform integrates deeply — non-technical employees drag and describe their way to agents: weekly-report summarizers, project-progress trackers, customer-feedback organizers. “Everyone has their own AI assistant” stops being a slogan.
Trace Work’s underlying logic: an enterprise’s core asset is not process, not relationships — it is knowledge. Whoever makes knowledge flow faster and get used more fully wins.
The Deep Logic: Leverage Points Determine Product Shape
| Dimension | Tencent WorkBuddy | Alibaba Tongyi | ByteDance Trace Work |
|---|---|---|---|
| Core leverage | Social relationship graph | Cloud + data infrastructure | Knowledge-flow efficiency |
| Entry scenario | Customer reach and conversion | Organizational productivity | Collaboration and knowledge management |
| Ecosystem base | WeChat / WeCom | DingTalk + Alibaba Cloud + e-commerce | Feishu + ByteDance recommendation engines |
| Delivery model | Enterprise custom + SaaS | SaaS + PaaS + cloud-native | SaaS + zero-code platform |
| Competitive edge | Unreplicable social network | Cloud + AI + data flywheel | Knowledge graph + algorithmic distribution |
The three paths are not mutually exclusive — a large enterprise can run WorkBuddy for customers, Tongyi for internal productivity, and Trace Work for knowledge simultaneously. But the divergence reveals deeper industry logic: a company’s core competence determines the angle from which its agent attacks the enterprise market. Tencent chose “use AI to amplify the value of connection.” Alibaba chose “use AI to activate infrastructure’s multiplier effect.” ByteDance chose “use AI to accelerate information flow” — from “people efficiently retrieve information” to “information actively reaches whoever needs it.”
The key insight: these three paths are not competing for the same market. Each locked onto a different bottleneck of enterprise digitalization — Tencent believes the bottleneck is customer-reach efficiency, Alibaba believes it is organizational productivity depth, ByteDance believes it is knowledge-flow congestion. Three bottleneck diagnoses, three product directions.
What the Frame Explains
The leverage-point frame resolves several seemingly contradictory industry facts. Why does Feishu penetrate SMEs faster? Because small teams’ bottleneck is information asymmetry, not process — Trace Work’s knowledge-first design fits teams that don’t need heavy approval flows. Why does WorkBuddy shine in retail and services? Because those industries live on the quantity and quality of customer relationships, and the WeChat-ecosystem connection is a direct moat. Why does Tongyi lead in manufacturing and government? Because those scenarios demand data security, compute control, and process compliance — Alibaba Cloud’s infrastructure depth answers them. And why doesn’t anyone copy anyone? Not inability — lack of incentive. Tencent cannot copy ByteDance’s recommendation engine and algorithm culture; ByteDance cannot copy Tencent’s social graph; Alibaba lacks ByteDance’s information-flow DNA. The paths diverged because three companies’ genes externalized themselves.
How to Choose
For managers evaluating enterprise agent platforms, the selection frame is direct. If your core business depends on building and maintaining customer relationships — retail, services, education, insurance — WorkBuddy’s social graph delivers the most direct ROI; your agent should work inside the WeChat ecosystem, not inside collaboration software. If your core challenge is internal process efficiency and data coordination — manufacturing, finance, government — Tongyi’s cloud-native architecture and process controls fit; you need a more efficient organizational operating system, not a smarter document. If your bottleneck is knowledge capture and information flow — tech companies, creative teams, consultancies — Trace Work’s knowledge-driven collaboration is the natural choice; your agent should solve “what should I know that I don’t.”
The worst choice is not picking the wrong platform — it is picking the wrong direction. Running customer relationships on Trace Work, or knowledge graphs on WorkBuddy, is feasible — and wastes each platform’s deepest leverage. The better strategy: identify your core bottleneck, then match the core leverage to it. Everything else is execution.
China’s three super agents are not really competing on model capability or feature count. Each company is exporting its core competence under the same “enterprise digitalization” headline, writing three completely different answers — and the grader is every enterprise that picks a platform.
