2025 Year-End Review: The AI Challenge Facing the SaaS Industry and the Survival Breakthrough for Listed Companies

Introduction: An Industry Reshuffle Accelerated by AI

By 2025, ChinaSaaSThe industry stands at a historic crossroads.

On one side, the retreat of capital has brought a rigorous scrutiny of profit models; on the other, generative AI has ignited a technological frenzy. While “AI+” has become the standard opening slide in every vendor's presentation, the market has already voted with its feet: customers no longer pay for flashy features like “chat capabilities,” but instead zero in on the core questions—“How much can it save?” and “How much incremental value can it deliver?”

According to the latest IDC data, China's enterprise-level SaaS market will reach 186 billion yuan by 2025, representing a year-on-year growth of 22.31%. However, this growth rate has slowed significantly compared to the 29.71% recorded in 2024. More concerning is that despite the AI SaaS sector being touted as experiencing a “full-scale boom,” its actual ARR (Annual Recurring Revenue) share remains below 15% of the overall market. A significant portion of so-called “AI features” remain confined to the demo stage, failing to translate into tangible business value.

“2025 will be the ‘AI stress test year’ for the SaaS industry,” Zheng Qingsheng, Partner at Sequoia Capital China, stated in an exclusive interview with this newspaper. “Many companies use AI as a fig leaf to mask core issues like product homogenization and sluggish growth. The true winners will be those that deeply embed AI into their business workflows, creating an irreplicable data flywheel.”

This article focuses on five representative listed companies: Beisen, Yonyou, Kingdee, Fanwei, and Zhiyuan (Note: “Jishuitan” is not a SaaS company and is likely a misattribution; Based on context, it may refer to healthcare IT firms like Weining Health, but due to its non-traditional SaaS model, it is excluded for now). Drawing on financial reports through Q3 2025, expert interviews, and frontline client feedback, it incisively analyzes their gains and losses amid the AI wave, ultimately addressing a fundamental question:Is AI empowering SaaS, or accelerating the elimination of pseudo-intelligent players?

I. Industry Panorama: Structural Crises Amid Rapid Advancement

1. Achievement: AI-Driven Paradigm Shift in Product Development

The most significant industry breakthrough in 2025 will be the fundamental evolution of SaaS from a “digital record-keeping system” to an “intelligent decision-making system.”

Taking HR SaaS as an example, Beisen's “AI Recruitment Agent” automates the entire process—from initial resume screening and interview scheduling to candidate profiling—reducing the average hiring cycle from 28 days to 17 days, boosting efficiency by nearly 40%. In the manufacturing sector, after integrating the AI scheduling engine into Yonyou U9 Cloud, an auto parts manufacturer saw its inventory turnover rate jump by 23%, significantly improving cash flow efficiency.

“In the past, SaaS addressed the fundamental question of ‘whether or not to digitize.‘ Now, AI tackles the core issues of ”usability, accuracy, and cost-effectiveness,“‘ noted Professor Wang Jianmin, Vice Dean of Tsinghua University's School of Software. ’By 2025, we will finally see the first wave of truly ”AI-Native' SaaS products come to fruition, marking the industry's entry into a new phase."

Policy efforts have also formed a concerted force. The “14th Five-Year Plan for Digital Economy Development” explicitly supports “industry-specific large models + vertical applications.” Cities like Shanghai and Beijing have established dedicated funds to encourage SaaS companies to collaborate with universities in building AI laboratories. This collaborative mechanism involving government, industry, academia, research, and application has accelerated the commercial deployment of AI in vertical scenarios such as finance and taxation, manufacturing, and human resources.

2. Failure: The Three Deadly Traps Beneath the AI Bubble

Beneath the surface of prosperity, however, dark currents are swirling, with numerous enterprises falling into the trap of “pseudo-innovation” in AI transformation:

Pitfall 1: AI Features Are Merely “Rebranded,” Lacking Core CapabilitiesAccording to iResearch's November 2025 “AI SaaS Commercialization White Paper,” over 601 SaaS vendors merely call OpenAI or domestic large model APIs to superficially package “intelligent Q&A” features, lacking deep training in vertical scenarios. One retail CRM vendor even labeled a customer service bot built on traditional rule engines and keyword matching as an "AI assistant," ultimately resulting in an 18% year-over-year decline in customer renewal rates and rapid market abandonment.

Trap Two: Mismatched Profit Models, R&D Costs Devouring ProfitsAI R&D requires continuous investment in computing power and talent, yet most companies still adhere to the traditional “subscription-based pricing model based on user numbers.” Kingdee's mid-year report for 2025 shows its cloud services gross margin at 61.21%. However, due to the lack of independent pricing capabilities for AI-related modules, these features are offered free to premium clients. This not only fails to boost ARPU (average revenue per user) but also drags down overall profit margins, creating a dilemma of “high investment, low returns.”

Pitfall Three: Organizational Capability Gap Leaves Transformation as an Empty Promise“Traditional SaaS companies excel at selling licenses and delivering projects, but AI SaaS requires the collaborative capabilities of three roles: data scientists, prompt engineers, and customer success managers,” admitted an industry insider who previously served as a senior executive at Salesforce China. “Many companies struggle to even hire AI product managers with business acumen, let alone implement AI solutions within customer scenarios.”

More critically, customer trust is eroding. Gartner's Q3 2025 survey reveals that 431 out of 400 enterprise CIOs explicitly stated they are “weary of SaaS vendors” AI hype,“ demanding suppliers provide quantifiable ROI (return on investment) proof—or face contract termination. The market has shifted from ”paying for concepts“ to ”paying for results."

II. In-Depth Review of Listed Companies: Who is Truly Innovating, and Who is Hyping Concepts?

clarificationAs of December 29, 2025, all A-share/H-share companies have yet to release their full-year 2025 annual reports. The following financial data is sourced from each company's disclosed 2025 third-quarter report (January–September) or interim report (January–June), with specific sources indicated in parentheses.

(1) Beisen (HKEX: 9680): The “AI Top Student” in HR SaaS”

What did we do right?

What did I do wrong?

Expert Commentary“Beisen is one of the few companies that has made AI the ‘business core‘ rather than an ”add-on plugin,“” said Zhang Fan, Research Director at Gartner China. "However, its high-end positioning has created a clear market ceiling. If it fails to rapidly penetrate the growing enterprise market, its long-term growth will face bottlenecks."

(2) Yonyou Network (SHSE: 600588): The giant elephant turns, but its steps are slow and laborious.

What did we do right?

What did I do wrong?

Financial AlertDespite cloud revenue exceeding 50% of total revenue, Yonyou's net profit for the first three quarters of 2025 declined by 9.31% year-on-year. This was primarily due to the contraction of its traditional license business (down 18.71% year-on-year) and the cloud business yet to achieve profitability. The capital market has already voted with its feet: the company's PE valuation has dropped from 45 times in 2023 to 22 times in 2025 (Wind data), reflecting market skepticism about the efficiency of its transformation.

(3) Kingdee International (HKEX: 0268): The Price of Aggression

What did we do right?

What did I do wrong?

Experts warn“Kingdee is repeating Oracle's mistakes from the 2010s—technologically advanced but with uncontrolled business momentum,” noted a Bain & Company partner specializing in enterprise software. “If it fails to monetize its AI modules independently by 2026, cash flow will remain under pressure and could even impact core business operations.”

(4) Fanwei Network (SHSE: 603039): The AI Dilemma Facing an OA Veteran

As the domestic leader in office automation, Fanwei attempted to break through with “AI-powered office solutions” in 2025, but achieved little success and found itself mired in a transformation dilemma.

Action Review

Fatal Weakness

Customer Feedback“Fanwei's AI is like putting GPS on a horse-drawn carriage—it's completely off track,” admitted the IT head of a financial group. “What we need are core AI capabilities for scenarios like smart contract review and compliance risk alerts, not peripheral functions like automatically generating meeting minutes.”

(5) Zhiyuan Internet (SHSE: 688369): The Low-Key, Pragmatic “Survivor”

Unlike peers who hype AI with flashy marketing, Zhiyuan adopted a strategy of “small, rapid steps and a focus on tangible results,” positioning itself as a survivor in an era of uncertainty.

Strategic Highlights

limitations

industry assessment“Zhiyuan is a classic example of a ‘cash cow” strategy. During industry consolidation, survival takes precedence over betting on the future,“ said an unnamed securities analyst. ‘However, its conservative approach may also miss out on AI-driven industry transformation opportunities. Moving forward, the company must strike a balance between ‘stability” and 'innovation.'"

III. AI's Fundamental Transformation of SaaS: Five Key Trends and Deep Insights

If 2023–2024 marked the exploratory phase where SaaS vendors “tested the waters with AI,” then 2025 will be the year of reckoning for “realizing value.” Based on in-depth observations of these enterprises, AI is reshaping the SaaS industry across five dimensions, prompting a series of profound strategic reflections:

1. Product Logic: From “Feature Overload” to “Intelligent Agent Collaboration”

Traditional SaaS centers on modular design, such as HRM comprising independent modules like recruitment, performance management, and compensation, with limited integration between modules. In contrast, AI-native SaaS adopts a multi-agent collaboration architecture as its core framework. Take Beisen as an example: after the Recruitment Agent screens candidates, it automatically triggers the Compensation Agent to generate offer recommendations. This then links to the Onboarding Agent to arrange training and activate IT permissions—the entire process requires no manual intervention, forming an “end-to-end intelligent closed loop.”

“The SaaS of the future isn't software—it's a ‘digital workforce team’ composed of multiple specialized agents,” stated a Silicon Valley AI expert who previously worked on large language model development. “This demands that SaaS vendors shift from traditional UI/UX design to agent behavior design and scenario collaboration design—a paradigm shift.”

2. Moat Migration: Data > Algorithms > Engineering

2025年行业最大的认知误区，是认为“接入大模型=拥有AI能力”。事实上，Universal Large Model只是基础设施，真正的竞争壁垒在于垂直场景的数据闭环。

“Yonyou possesses tens of millions of financial voucher data points, while Beisen holds hundreds of millions of talent behavior trajectory records. These datasets, accumulated over a decade, have formed an irreplaceable ‘industry knowledge graph,’” emphasized Huang Mingming, Founding Partner of MingShi Capital. “Without such vertical data, your AI remains merely an ‘echo’ of others” large models, incapable of solving customers' real-world problems."

This also explains why startups struggle to break through in mature sectors like HR and ERP—it's not that their technology is inadequate, but rather that leading players have already locked in the data flywheel, making it difficult for latecomers to overcome.

3. The Revenue Model Revolution: The Rise of Performance-Based Payments and the Challenges Facing Subscription Models

Traditional SaaS models charge based on user count and module count, but AI's core value lies in “outcome delivery,” not “feature usage.” By 2025, a wave of pioneering enterprises began experimenting with novel pricing models:

“Subscription models are being challenged by ‘value-sharing systems,”“ McKinsey noted in its report ”Trends in AI Commercialization in China by 2025." "Customers are willing to pay for guaranteed outcomes rather than potential features—this will fundamentally reshape the profit logic of SaaS."

4. Customer Success Evolution: From CSM to “AI Usage Coach”

The core responsibilities of a traditional Customer Success Manager (CSM) include training, troubleshooting, and driving upsells. However, in the era of AI SaaS, CSMs must evolve into “AI Usage Coaches”—teaching clients how to design prompts, calibrate model biases, and translate AI recommendations into tangible business actions.

Kingdee piloted the “AI CSM” role in Q3 2025. Through specialized training, the number of clients served per person increased threefold, while client AI feature usage rates rose from 32% to 68%. However, the challenge lies in the extreme scarcity of such hybrid talents possessing both business acumen and AI expertise, creating a widespread bottleneck across the industry.

5. Ecological competition replaces individual efforts; integration capabilities determine success.

No single company can master computing power, models, and scenarios on its own—by 2025, ecosystem collaboration will become the standard for AI SaaS:

“The future SaaS war will not be fought over individual products, but over ecosystems,” concluded Zheng Qingsheng of Sequoia Capital China. “Whoever can integrate the best models, the deepest scenarios, and the most stable computing power will define the next generation of intelligent service standards.”

IV. The Ultimate Test for AI SaaS

To deepen industry insights, we have synthesized perspectives from multiple authoritative experts to distill three core judgments:

Judgment 1: AI will not eliminate SaaS, but it will eliminate non-intelligent, non-native SaaS.“In the next five years, all SaaS solutions will either become AI Native or be phased out,” asserted Professor Wang Jianmin of Tsinghua University. “By AI Native, we don't mean simply adding a chatbox—it means building products from day one around..."AI AgentDesign, data closed-loop, intelligent collaboration, quantifiable value—this is an irreversible trend.”

Judgment 2: The breakthrough point for China's SaaS lies in “Industry Know-How × AI”“U.S. SaaS wins through standardization, while China must rely on deep industry customization,” Bain & Company's technology practice noted in a recent report. “Manufacturing production planning logic, hospital treatment pathways, government document workflows—these years of accumulated industry know-how represent rich mines for AI implementation and form barriers foreign companies struggle to overcome.”

Judgment Three: Regulation Will Be a Double-Edged Sword for AI SaaS““The Interim Measures for the Administration of Generative AI Services mandate algorithmic explainability and data traceability, imposing stricter requirements on AI SaaS solutions. A leading law firm's technology compliance partner cautioned that SaaS offerings in sensitive sectors like healthcare, finance, and government services risk being taken offline if they cannot provide clear justification for their recommendations. However, compliance will also serve as an industry barrier, weeding out non-compliant small vendors.”

V. Conclusion: 2026, the knockout stage officially begins.

By 2025, the SaaS industry had completed its “mass experiment” with AI, yielding a brutal and clear outcome:

“Over the next three years, China's SaaS market will transition from a hundred schools of thought vying for dominance to a fierce battle among ten major players,” Zheng Qingsheng predicted. “AI will no longer be a bonus feature but a matter of survival. Companies unable to demonstrate that AI delivers quantifiable business value will be completely kicked off the table.”

For investors, the logic for stock selection in 2026 has fundamentally shifted: the focus is no longer on “whether a company has integrated AI,” but rather on “whether AI drives NDR growth and establishes independent monetization pathways.” For entrepreneurs, instead of chasing the large-model trend, it's wiser to deeply cultivate a vertical scenario and build a “small yet beautiful” intelligent closed-loop system.

This AI-driven test has only just entered its second round. Those caught with their pants down will eventually be exposed when the tide recedes, while the true era of native intelligence is accelerating its birth through a brutal elimination process.