2026 Global Digital Ecosystem Summit

The WorkBuddy enterprise productivity AI agent has been the big story for Tencent this year. Launched in March, the product has proved hugely popular in China, becoming the country’s most widely used AI productivity tool. The aim of the platform is to shift traditional workflows to AI-co-ordinated tasks to improve enterprise productivity, says the company.

Despite this, Tencent’s share price has been in steady decline this year, perhaps reflecting an AI market that is cooling in terms of public fervour, even as it is becoming more established in the enterprise. 

In general terms, Tencent – in common with other Chinese majors – is also facing a scarcity of compute and is still reliant to a degree on US GPU providers, including NVIDIA and Oracle, as I explored in my earlier report. All this may be a worry for investors.

But at Tencent’s 2026 Global Digital Infrastructure Summit in Shenzhen, WorkBuddy was front and center of that infrastructure message from the conference stage, while a huge booth for the platform was given pride of place in the exhibition hall. 

Real-world mission

Jade Ding is Head of Overseas for WorkBuddy. She explained that the platform is live in over 21 markets, with targets including Hong Kong, Macau, Singapore, and Southeast Asia. The platform focuses on solving real-world tasks, she said, such as data analysis and coding, rather than focusing on “persona discrepancies” between markets.

Addressing the challenge of users’ legacy systems, Ding explained that the ability to “dock with” AI depends on a company’s internal data infrastructure, which makes the fast-emerging role of FDE (Forward Deployed Engineer) critical in how well legacy data can be integrated with AI-native systems.

Speaking in Mandarin, with in-person translation provided by an interpreter, Ding said:

Integrating and docking with multiple models, including ChatGPT, Grok, Gemini, Hy, and DeepSeek. So those docked or integrated models are popular and influential around the world, and we can offer them combined with multi-model agent solutions. 

Second, the product positioning of our AI agent is always focused on solving real-life tasks in production, in production workspaces, and improving their productivity. Our goal here is always to help users to transfer the traditional workflow into an AI-coordinated mode that could utilize, and benefit from, the experience of AI capabilities. 

So, I think with these two positionings, we are poised to compete in the international market. Although, as an experienced professional in AI field, we often see headlines like AI is picking up very quickly. But in real life, in ordinary lives and workspaces, the penetration rate of AI agent is still extremely low. So that really sheds light on the future potential of a big market opportunity.

Enterprise perspective

As to whether most so-called enterprise AI adoption is still individuals’ shadow use of the technology within organizations, or whether it has shifted towards strategic adoption by enterprises themselves, Ding said:

I can’t give a specific answer, but I can say that this phenomenon is quite common in the international arena right now. From an enterprise perspective, normally when advanced technology emerges, it is always those core technology guys who will make a trial and use it and see if the productivity can be boosted dramatically. If so, they will promote it to a team and then expand it to the whole organization.

There is a new emerging job called FDE. Their responsibility is to delve deep into the internal structure of the company and evaluate if the legacy data infrastructure can be docked to the AI-native system, and how big the gap is. So, some companies can easily dock with AI native without much fixing, but other companies will need to make more efforts in the fixing of their old systems.

Ding’s argument is that AI transfers ‘dirty work’ to machines while leaving creativity and decision-making to humans – a message that has been the mantra of automation providers for much of this century. In the real world, however, many users have deployed AI to automate creativity rather than solve critical business problems.

That aside, a growing concern among enterprise users this year has been the security of AI systems – especially agentic AI, given the many high-profile examples of rogue agents breaking into external websites in recent weeks. WorkBuddy needs to be able to address those security fears, given that it allows people to connect and trial disparate models for different tasks. Ding explained: 

I think it’s a focus of all AI platforms right now. WorkBuddy has always prioritized data security in terms of integrating security and placing guardrails into the architectural design of our products and functions. So, we have placed inherent guardrails in the function design, and in the architectural design. So those are already embedded in both architecture and the product systems.

On WorkBuddy’s day-to-day operation, she said:

A lot of that work involves a lot of redundancy of documents, files, and information, so you can feed all of this redundancy of information into WorkBuddy, and it can call or invoke the appropriate models to analyse the prompts, the intentions, or the context of the questions that you feed into it. And it calls for appropriate tools to solve the task and deliver the final output or outcome. 

So, you can see that WorkBuddy understands deeply the context and intention behind the directives, autonomously calls or invokes tools, and offers complete delivery of the final output. Other than that, delivery can be optimized continuously based on user habits, requirements, and the memory of previous tasks’ solutions.

My take

Make no mistake, WorkBuddy is positioned as Tencent’s route to discreet global expansion, offering users choice – including to switch away from their US model for some tasks and explore other options, via open-weight Chinese alternatives.