Tencent WorkBuddy Turns Office Agents Into a 'Cyber Lego' for Individuals and Teams

The report included ease of use, collaboration and long-term use in its evaluation metrics, the review said. The two lists reflect two demands that office tools eventually face. New users want results today, while teams care about shared methods, permissions and correction. WorkBuddy addresses both through expert roles, file editing and team spaces, according to the review.

On first opening, WorkBuddy offers tasks, assistants, projects, experts, automation, a data library and Skills. Its Expert, Skill and Connector plaza organizes capabilities by office scenarios such as data analysis, recruiting, marketing and content. In tests, the review selected a Tech Channel Topic Evaluator for an iPhone DUO topic and asked for angles aimed at hardcore geeks, digital enthusiasts and general readers; WorkBuddy checked the release date and related facts before scoring. It also uploaded fictional robot vacuum Q3 sales data and asked a data analysis expert to identify the two brands with the largest month-on-month decline, calculate average selling price changes and explain whether discounts drove sales, producing bar charts and analysis. Other tests included a live commerce coach creating a 45-second script for a magnetic cooling power bank and a senior marketing planner arranging an AI PC salon under a 200,000 yuan budget, 10 media outlets and three hours.

Because AI can operate files, the review asked whether originals can be recovered. WorkBuddy 5.0 and later versions have a Security Center with default-on features: real-time interception of unauthorized operations, approval for sensitive operations, automatic movement of deleted files to the recycle bin and file version restoration. The review said this recoverability affects whether users will hand real work to an agent.

The review said it often uses WorkBuddy's human-AI co-writing function, which supports Tencent Docs and local Word, Excel, PowerPoint and Markdown files. In a simulated product launch draft, it circled an eco-packaging paragraph and asked WorkBuddy to keep the biodegradable and 30 percent carbon reduction information while changing only wording; changes synced to the original file and other content stayed unchanged. In a business review report, it manually updated Word figures and asked WorkBuddy to supplement paragraphs; the agent used the latest numbers, modified only specified locations and kept the format. In PowerPoint, it preserved bullet points and page structure while rewriting selected text. In a cross-format task, it asked WorkBuddy to extract a compliance document chapter on data security and asset protection into a new employee training presentation outline and save it to the data library; the agent produced a logical outline, then generated a presentation and adjusted wording through co-writing. The review said the user decides the scope and final wording while AI processes the latest file version.

The review also tested WorkBuddy as an organization operating system, connecting lead finding, topic selection, writing, review and knowledge accumulation. It said many companies see limited company-wide efficiency gains even when one person uses AI well. WorkBuddy's data library builds editorial memory by organizing historical articles, long-term topic materials and confirmed writing preferences. Its updated team space lets multiple people maintain content. Personal documents can be moved into team space and classified by project and directory; interview materials, topic records and agent outputs can sit together. Senior editors curate and write, new staff call on verified materials, and managers control members and permissions. The agent can only access content the current user is authorized to view, according to the review. A new hire can start from selected accumulation of more than ten years.

For human-human-machine collaboration, the review described a recent Mac mini project requiring multiple people. An editor shared a draft through the data library, and the editor-in-chief and testing colleagues annotated the original text. Colleagues then let their own agents read the annotations and propose changes; when uncertain, the agent asked a person before writing back. The review said it could see who initiated changes, which suggestions were accepted and previous versions, with opinions and reasons left together. After each collaboration, the agent can distill revision comments into a skill, bringing its editing style closer to the person. WorkBuddy also supports custom models for different stages.

The review cited Frost & Sullivan data showing current use of cross-system data collaboration is only 14.3 percent, while enterprise users value its long-term value. Many work processes remain in scattered documents, spreadsheets and chat records. In the review's view, office agents will move toward letting people who know the business participate directly in tool design, writing requirements into workflows for colleagues to reuse. WorkBuddy aims to be a cyber Lego for office workers, allowing methods to be shared and recombined, starting from a weekly report or a team-built work system.