Tencent is extending its WorkBuddy artificial-intelligence platform from productivity assistance into operational enterprise systems. Through Model Context Protocol connections, agents can retrieve information from enterprise resource planning, customer relationship management and other applications, combine records, generate reports and, where authorized, write actions back. The design aims to replace isolated chat prompts with workflows that cross multiple systems.
A practical example is expense processing. An agent could read policy and employee data, inspect a reimbursement request, prepare an analysis and submit or batch an approved transaction. The same pattern can support sales summaries, service cases or internal research. Reading and writing are different risk levels, however: a wrong answer in a report can be reviewed, while an incorrect update can alter financial or customer records.
The broader Buddy One package brings together Tencent Docs, Meeting, knowledge, design and other services under one account with shared credits and data. Tencent Drive supplies memory for context and output. WorkBuddy can also appear in the WeCom sidebar, where administrators allocate or reclaim seats, tokens and permissions, placing the agent in tools employees already use rather than in a separate destination.
Tencent is keeping the architecture relatively open. Companies can connect third-party or internally built agents, use external messaging tools and select among models. TokenHub offers common billing and a single interface for Tencent’s Hunyuan models and alternatives including DeepSeek, GLM, MiniMax and Kimi. That choice can reduce integration duplication, but it also requires model routing, data-policy and vendor-risk rules.
The company cites Yili as an operating example: more than 1,000 sales advisers use a combination of WorkBuddy, WeCom and an enterprise personnel system. Tencent says operational efficiency improved 50% and that an A/B test produced 26% more sales orders for the agent-assisted group. Those are vendor-reported results from one deployment, not an independently audited productivity benchmark that can be generalized to every company.
Tencent also says WorkBuddy has handled more than nine billion task requests since March, covers more than 50 industries, is used by half of China’s top 500 companies and has attracted 50,000 developers and partners who created 200,000 skills. These figures indicate platform reach under Tencent’s definitions, but they do not reveal active paid users, revenue, retention, task complexity or the proportion of requests completed without human correction.
Governance becomes more important as agents gain write access. Enterprises need least-privilege identity, approval thresholds, logs, segregation of duties, protection against malicious instructions and controls over which data each model receives. Legacy ERP and CRM systems may expose inconsistent interfaces. Token consumption can rise rapidly in multi-step tasks, so companies must match model quality to task value and monitor cost per successful outcome.
The commercial test is whether WorkBuddy can make cross-system work reliably cheaper and faster without creating new operational or compliance failures. Adoption claims will be more persuasive when accompanied by customer renewal, error rates, latency, audited savings and revenue disclosures. Connecting agents to enterprise systems is technically consequential, but durable value depends on controlled execution rather than the number of prompts processed.
