At the World Artificial Intelligence Conference (WAIC), a shift in focus was evident compared to previous years, which centered on technical metrics like parameter size and multimodal capabilities. This year, the conference signaled a return of AI from being a "technological marvel" to an "essential tool".

As AI collaborates in automobile assembly, multimodal models accurately diagnose diseases, and AI reshapes financial investment research processes, the discussion around AI is evolving. It's moving away from labor substitution, technological competitions, and functional stacking, towards practicality, commercialization, and ecological integration, pursuing genuine efficiency and usefulness. This transformation reflects a renewed understanding of AI's essence across various industries.

The office sector is particularly illustrative, where the core scenario contradiction is always "efficiency versus cost." Writing reports, creating spreadsheets, editing PPTs—these tasks are repetitive, trivial, yet indispensable. Being a "high-frequency rigid demand" means the value of AI here doesn't require market education. Users' judgment of "usefulness" is straightforward: whether it saves a mouse click, a format adjustment, or a sleepless night.

Taking Kingsoft Office's latest release at WAIC as an example, unlike last year's WPS AI Enterprise and Government Editions targeting B-end scenarios, which enhanced organizational efficiency through AI Hub, AI Docs, and Copilot Pro, this year's focus is more on the C-end.

The launch of WPS Lingxi announces that WPS AI has officially entered the 3.0 era. Users can complete functions like document creation, presentation generation, and voice assistant through natural language and multi-round dialogues, without complex operations or external jumps throughout the process. It's a genuinely native Office intelligent agent rooted in user needs.

As a 37-year-old company, from setting AI as one of its strategic goals in 2018 to the current release of WPS Lingxi, Kingsoft Office has gradually built a comprehensive AI office ecosystem.

This is what intrigues us the most—a company with 37 years in the office sector can break consensus and redefine the industry through continuous innovation. Upon deeper exploration, we found counterintuitive aspects to the company's recent development, revealing a new path for the rise of domestic office software.

01

The "Bidirectional Integration" of AI and Office Software: More Than Just "Functional Stacking"?

When WPS AI 1.0 was released in 2023, it still carried industry inertia. The product logic was simple, influenced by Microsoft Office-like products, mainly focused on "quickly implementing AI functions," treating AI as a functional plugin to be added to existing software.

Tian Ran, Assistant President of Kingsoft Office, admitted frankly in an interview, "At that time, we still had inertia thinking, believing that AI should be like 'generating a report or intelligent summary with one sentence'."

However, with the continuous accumulation of user data, behind the 100 million daily active users and millions of AI function calls lies a myriad of "not flashy but painful" needs: people fearing formatting errors when modifying contracts, making mistakes when calculating data with spreadsheets, and not knowing how to streamline dozens of pages of PPTs.

These needs highlight a truth overlooked by the industry: the core of AI in office scenarios is not "creating content" but "solving specific problems".

This shift in cognition directly gave birth to the core logic of WPS AI 3.0 – "integration of users, software, and AI".

Previously, product improvements were mostly technology-oriented incremental adjustments: treating software as a container and AI as a plugin, cramming various AI functions into the framework of existing software functions, and requiring users to learn prompts and adapt to AI interactions.

In contrast, WPS AI 3.0 focuses on users' original needs, no longer requiring users to accommodate software or AI. It breaks the fixed boundary between software and AI, disassembles software functions and AI capabilities, and then recombines them using their respective strengths. As Tian Ran said, "Forcibly cramming AI into software or making software accommodate AI is not the optimal solution. The key is to let users, software, and AI grow together around real needs."

A typical achievement of this stage is the implementation of the "edit while chatting" interaction mode. For instance, when users modify contracts, in the past, AI directly generated a new contract that couldn't be used directly. Compared to such flashy approaches, WPS AI 3.0's "edit while chatting with unchanged format" is clearly more practical.

In this process, software reconstructs underlying capabilities for AI, and AI also learns to "communicate well" with software—software ensures precise formatting and reliable operation, while AI can understand users' natural language and directly make modifications in the original document. The generated content must be usable, ultimately serving the original need of "users spending less time modifying contracts".

Based on this underlying logic, the native Office intelligent agent "WPS Lingxi" emerged. As a genuinely native Office intelligent agent, it also marks a complete reconstruction of traditional office AI thinking.

Office scenarios have never been isolated "tools" but rather an ecosystem comprising "documents, collaboration, data, and devices": a contract may need to be drafted on a computer, annotated on a mobile phone, and archived in an enterprise system; a PPT may need to interface with cloud storage, meeting software, or even smart projectors. The complex ecological attribute determines the integration route of AI here.

Taking WPS Lingxi as an example, it's not an independent AI tool or software but, as Tian Ran puts it, "a service," an intelligent agent embedded in WPS Office that can directly invoke spreadsheet formulas, PPT layout systems, and cloud document permission systems. This "bidirectional transformation of AI and software" is precisely the natural requirement of office scenarios for "ecological collaboration".

Users can complete functions like document creation, presentation generation, and voice assistant through natural language and multi-round dialogues, without complex operations or external jumps throughout the process. It's not "AI for AI's sake" nor does it become a "super entry" by piling up functions. Instead, it solves almost all office-related problems within the entire ecosystem while providing basic office capabilities, accessible to everyone.

Many may not realize the significance of this.

According to official data from Kingsoft Office, as of the end of March 2025, the global monthly active device count for WPS Office has reached 647 million. In other words, WPS Lingxi can reach a massive user base upon its release, which is likely to become a landmark event in the history of office intelligent agents.

02

Behind the Low-Threshold Interaction: Balancing "Efficiency Improvement" and "User Independent Thinking"

"Natural language dialogue, no prompts required," WPS Lingxi lowers the barrier to entry for AI office use, but this also raises a common concern: as AI becomes more adept at understanding users, will it weaken users' ability to think independently?

Through dialogue with the Kingsoft Office team, we found that WPS AI has always clearly opposed "AI replacing humans." On the contrary, they emphasize that "AI will further amplify users' creativity."

Tian Ran used the analogy of "mathematical tools," where the advent of calculators freed humans from the need to practice manual calculations, but mathematical thinking and modeling abilities became more valued. Similarly, AI raises the "baseline level" of office work, allowing ordinary people to quickly generate standardized documents, but the value above 95 points still comes from users' judgment, discrimination, and deep thinking.

WPS Lingxi's "edit while chatting" interaction mode is a key design. After users propose their needs, the content generated by AI is not the "final draft" but requires continuous feedback from users. This design avoids the passive acceptance of "AI generation is the end," keeping users in a state of "decision-making, adjustment, and optimization" throughout the process.

Lowering the barrier to entry and amplifying individual differences, from "general tools" to "personalized intelligent assistants," this is the core evolutionary logic of WPS AI.

You'll find that WPS Lingxi's application in AI writing and AIPPT functions almost perfectly solves the problem of traditional AI tools being easy to generate but difficult to edit:

The former adopts a "comparison editing" mode between the "original text" and the "revised version," supports adjustment via natural language instructions, and can accurately recognize complex symbols; the latter first refines needs through outline collaboration, and after generation, all text and images can be directly edited, supporting changes to layout, color scheme, animations, etc. It not only generates content in real-time but also allows text box-level editing without format loss.

Moreover, as the first knowledge base product in China based on native Office documents, unlike previous knowledge bases that required users to proactively collect and organize information, WPS Lingxi provides each user with a dedicated information assistant: by integrating various documents in the user's cloud storage, using self-developed algorithms to achieve cross-format parsing and semantic association, and then letting AI generate content based on the user's personalized materials rather than relying on general data.

Imagine needing to write an annual work summary but with project documents, weekly reports, and customer feedback from the past year scattered in your cloud storage. You only need to ask WPS Lingxi to automatically search the document library and extract key information, which it will return with annotated sources through semantic understanding, rather than just returning files.

In this way, not only is creativity traceable, but it also avoids starting from scratch.

Additionally, the newly launched Lingxi voice assistant is also a highlight, supporting dialect and colloquial voice commands on mobile devices, quickly extracting document essentials, and generating a preliminary framework during conversations, transforming "fragmented scenarios" into creative incubation time.

A typical scenario is commuting when you suddenly remember you need to add market size data for the next day's meeting but don't have a computer at hand. Open WPS Lingxi and say, "Take a look at the 'XX Industry Report' in my cloud storage and extract relevant data, such as annual growth rate and the proportion of leading enterprises." WPS Lingxi will quickly return structured essentials and voice the key data.

The clever design of this function, on the one hand, breaks scenario limitations to capture inspiration, allowing for quick information processing in fragmented scenarios like commuting and meeting breaks through voice interaction alone, avoiding the loss of inspiration; on the other hand, it simplifies operations to keep thinking coherent, eliminating the need to manually flip pages to find data or type frameworks, allowing users to focus on specific document information needs and making the transition from "vague ideas" to implementation smoother.

Whether it's the edit-while-chatting of AIPPT, the personalized integration of the knowledge base, or the fragment capture of the voice assistant, WPS Lingxi allows AI to handle "standardizable mechanical work" like format adjustment, information retrieval, and basic framework construction, leaving "irreplaceable human wisdom" like judgment, insight, and creative conception to users. As Wang Yingren, head of WPS mobile products, said, AI brings not replacement but collaboration; AI is not the end of reading but the beginning of understanding. On mobile devices, WPS Lingxi is initiating a new way of dialogue between humans and information.

This design of AI assisting rather than replacing is the key to amplifying creativity.

03

Defining "Useful": The Key Indicator for Measuring the Value of AI Office

From bamboo slips to electronic files, documents as the core medium for human information communication will retain their value for a long time.

When asked whether WPS will become a super entry, Tian Ran's answer was thought-provoking: "If office software becomes a super entry, it means everyone has no life." In his view, the ultimate form of WPS AI is not "doing everything" but "doing one thing well"—Office capabilities will become essential skills for every intelligent assistant, like water and electricity.

This positioning is reflected in the evolution of product forms, with WPS Lingxi transforming from a "sidebar tool" to an "embeddable capability." In his view, future AI office tools can even be embedded in smart glasses to annotate PPTs in real-time during meetings; they can be integrated into automotive systems to voice-edit weekly reports during commutes; and they will also be applied in smart homes to convert shopping lists on refrigerators into Excel spreadsheets.

This naturally brings more commercialization opportunities to the AI office industry.

The business logic of the office sector is clear: users are willing to pay for "quantifiable efficiency improvements" rather than for AI technology itself. However, the industry often assumes that "the more useful a function is, the more willing users are to pay," but there's always a lack of a user perspective on how to define "useful".

Tian Ran believes that "many AI products today do not talk about charging because they are still in the stage of verifying user needs. When they find the entrance to a positive cycle and establish a basic user base, charging becomes the golden principle for judgment—things that users are unwilling to pay for are essentially not well-done."

The WPS Lingxi team prioritizes "addressing user pain points before discussing monetization," viewing charging as a testament to validated value, rather than merely a profit-making tool. "Users are prepared to pay for solutions that genuinely address their issues. However, if a feature is merely flashy, it won't attract payment, regardless of its price."

Therefore, in contrast to the historical tension between the often-overlooked "value verification cycle" and "commercialization rhythm," WPS AI has opted for a path of iterative demand accumulation on the consumer end (C-end) and value precipitation in high-value scenarios on the business end (B-end). This approach serves as a response to the aforementioned contradiction.

Reflecting on WPS AI's evolution over the past two years, from launching the Enterprise Edition to create an "enterprise brain" for clients to introducing WPS Lingxi aimed at C-end users, the willingness of C-end users to adopt it underscores its genuine value, while B-end users' readiness to pay indicates the viability of commercialization.

Currently, this "two-legged walking" model has been market-validated, evidenced by double growth in user scale and commercial value as reflected in financial reports:

In 2024, the WPS 365 business debuted as an independent item in the report, experiencing a remarkable 149% year-on-year surge to reach 437 million yuan. This indicates a growing shift among clients from using WPS software to opting for WPS 365 services. In terms of individual business, the cumulative monthly active device count surpassed 600 million, with both revenue and gross margin on the rise. This year's first quarter saw continued steady growth in revenue and scale for WPS 365 and individual business.

Fundamentally, the core of office scenarios revolves around "people," and the value of AI here lies not in replacement but in empowering individuals and organizations to achieve value jumps with lower costs and higher efficiency. In the intensely competitive AI office landscape of 2025, Kingsoft Office's absolute advantage hinges on its unparalleled "product-data-ecology" closed loop:

Without 37 years of foundational Office expertise, it would be impossible to "generate native format content with AI." Without genuine data from 100 million daily active users, AI iterations would stray from practical scenarios. Without an ecosystem and security system spanning individuals to enterprises, service orientation would remain a mere concept.

This deep accumulation not only positions WPS uniquely in the integration of AI and office software but also anchors its vision for the future of the industry, a feat that WPS executes best.