How an AI‑Native Office Assistant Can Cure Work‑Place Anxiety

By 2026, AI agents are converging on a common capability list—search, generation, computer operation, and code writing—yet most AI‑generated office outputs remain "pseudo‑documents" that require extensive manual rework.

At the WAIC‑hosted "2026 AI Productivity Conference" in Shanghai, Kingsoft Office unveiled its independent AI‑native office agent, Lingxi Professional Edition, positioning it as a fundamental shift: "Not adding AI to Office, but letting AI grow Office." This marks what the author calls the most thorough logical turn in office software in four decades.

Core Logic and Project‑Based Memory

Lingxi organizes work around projects . All conversations, documents, data, and progress for a given project are stored in a unified context, allowing the agent to retrieve background information automatically for any new dialogue—effectively giving the AI a continuous onboarding experience.

Multiple domestic large models are integrated; the system automatically selects the most suitable model for each task without user intervention.

Delivery Capabilities

When tasked with writing a report on the "2026 AI Productivity Conference," Lingxi autonomously searched news, read core articles, and produced an outline. After providing additional constraints (focus, audience, length), it generated a fully‑formatted .docx file with proper heading hierarchy and styles, ready for revision.

For a 5,000‑word research report on the Kimi K3 model, the agent produced the text and then created a 15‑page PPT. During generation it asked for a visual style, built an outline, rendered each slide, performed layout checks, and corrected issues, delivering an 18‑page presentation with editable native elements.

Engineering Chain and Native Document Generation

Lingxi is a standalone AI agent that collaborates with WPS rather than being embedded within it. It leverages the WPS CLI , an open document‑operation kernel built on decades of WPS engineering, to create genuine Office‑native objects. Generated PPTs contain independent image elements, editable tables, native charts with data sources, and modifiable text placeholders, all conforming to Office standards.

In spreadsheet scenarios, Lingxi preserves formulas, conditional formatting, and data relationships, enabling automatic updates when source data changes—unlike typical AI‑generated static tables.

Before delivery, Lingxi runs a rendering validation step that detects layout errors, overlapping elements, or formatting issues and automatically corrects them, ensuring high‑quality output.

Collaboration and "Automatic Dreaming" Memory

Users can share a project link to invite teammates into the same AI‑driven workflow. The newly added "automatic dreaming" feature runs during idle compute periods, summarizing daily conversations and files, compressing and structuring information to prevent context loss in long‑running tasks.

The memory system has three layers: (1) user‑defined personal info, writing style, and collaboration rules; (2) uploaded project materials forming a knowledge base; (3) automatically generated summaries from the dreaming process. All memory is user‑controlled, visible, editable, and can be disabled.

AI Office Evolution and Market Landscape

The author outlines three AI‑office eras: 1.0 – conversational AI that answers questions and generates content but does not act; 2.0 – Copilot‑style plugins that enhance existing software without changing the underlying workflow; 3.0 – AI agents that autonomously decompose goals, invoke tools, and deliver usable results, shifting the human role to decision‑making and acceptance.

Globally, major players (Anthropic’s Claude Cowork, Tencent’s Workbuddy, OpenAI’s ChatGPT Work) have launched AI agents, driving rapid user adoption. However, the industry faces a common challenge: while agents can operate, delivering truly usable outputs remains difficult.

Kingsoft Office argues that most AI office products output static documents lacking editable formulas or charts, requiring extensive rework. Lingxi’s advantage lies in its deep integration with WPS’s document engine, a capability that generic large‑model providers cannot replicate quickly.

Strategic Perspective

Kingsoft CEO Zhang Qingyuan frames AI office as the third era of office software, moving from manual operation to goal‑driven AI collaboration. The company believes that as large‑model capabilities become commoditized, competitive advantage will stem from deep contextual understanding and accumulated office data assets.

Assistant President Tian Ran defines three hard standards for AI office assistants: (1) understanding user context, (2) high‑quality task completion, and (3) professional Office operation skills. Lingxi meets these through multi‑model orchestration, project‑based memory, and native WPS document generation.

In summary, Lingxi Professional Edition demonstrates that an AI‑native office agent can truly advance from merely generating drafts to delivering complete, editable, and production‑ready work artifacts, leveraging decades of WPS engineering to overcome the engineering gap that hinders most AI office solutions.