WorkBuddy's Bet Isn't on the Office Entry Point

On September 2nd, Tencent pushed the WorkBuddy open platform into a three-layer ecosystem of hardware, applications, and developers. At first glance, it looks like another AI office entry point; but what I care more about is whether it can transform agents from showcase capabilities into a capability layer that can be stably called by business systems. The entry point is the surface layer, the ecosystem is the middle layer, and capital expenditure is the trump card. This quarter, Tencent's capex was in the range of over 50 billion yuan, nearly doubling year-on-year.

Having just gotten to know agents like WorkBuddy for less than a week, my feeling is clear: model capability is just the ticket to entry. The value of the platform depends on whether the interfaces are stable, permissions are controllable, and whether developers are willing to integrate long-term. Compared with overseas Copilot and Agentforce, the direction is the same—embedding into business systems. AI won't bypass SaaS directly; instead, it will break SaaS down into backend capabilities.

An open ecosystem sounds smooth, but developers need to first believe that this set of interfaces isn't just an accessory feature of their own apps. If hardware, applications, and interfaces can all be stably called, then WorkBuddy is an enterprise workflow runtime; if they still operate in silos across devices, it ultimately just moves users from multiple interfaces into yet another interface.

Whether WorkBuddy succeeds depends on whether enterprises are willing to hand over process permissions to a platform-level agent.