By Lao Yu'er, Edited by Yang Xuran
Recently, a company that has not received massive attention but is also a key player in the important application of AI (Artificial Intelligence) is sprinting towards the capital market. Zhuhai KingSware Artificial Intelligence Co., Ltd. has submitted a new listing application to the Main Board of the Stock Exchange of Hong Kong (HKEX). This marks its second attempt after the initial filing on December 15, 2025, lapsed upon reaching the six-month validity period.
The prospectus shows that, in terms of the number of licenses, by the end of June 2026, KingSware had deployed over 2 million AI digital employees, "engaging in more than ten industries including financial services, government affairs, and manufacturing, cumulatively serving over 1,600 clients (including more than 120 Fortune Global 500 and Fortune China 500 companies)," and "providing services to over 270 banks (covering China's six major state-owned banks), over 130 securities companies (covering more than 90% of securities companies in China), and over 220 other major financial institutions."
Observing this company that produces digital employees from a business perspective, its operations are future-oriented with prominent technological attributes, which should theoretically yield strong financial performance, capital favor, and regulatory support. However, the financial data tells a more complicated story.
From 2023 to 2025, the company accumulated a total loss of approximately CNY 531 million over the three years. Its cash and cash equivalents on hand dwindled to merely about CNY 72 million, while liabilities related to redeemable preferred shares reached as high as CNY 2.126 billion.
A company that has produced 2 million digital employees and a technology enterprise with a stellar client roster ultimately failed to achieve sound financial performance. Behind this lies a story about AI, business models, and operational validation.
This article is an in-depth value piece from the content team of Juchao WAVE. You are welcome to follow us across multiple platforms.
Appeal
The story of digital employees, a niche track in AI, certainly sounds appealing enough.
A report released by iFenxi in July 2026 estimates that the market size of digital employees in China will surge from CNY 7.47 billion in 2025 to CNY 209.84 billion in 2030, with a compound annual growth rate (CAGR) of up to 94.9%. The underlying logic behind this explosion is attributed by the institution to "the qualitative leap of digital employees from cost-reduction tools to a new type of workforce."
In its prospectus, KingSware also defines AI digital employees as "a virtual workforce integrating RPA and AI technologies (such as OCR, ASR, NLP, and Large Language Models), capable of replacing humans to complete rule-based, repetitive, and cross-system operational tasks in enterprises and government departments."
Once digital employees are defined as "workforce," their connotation and extension undergo a comprehensive upgrade. The ceiling becomes higher, and the necessity for institutions to make relevant arrangements increases significantly.
In the past, budgets for information technology and the like were categorized under corporate IT informatization expenditures. However, with the emergence of the "human resource attribute" of information technology, its track expands from IT information to human resources.
This logic is undoubtedly quite attractive: if digital employees, as a commodity, can truly be incorporated into the corporate headcount just like "humans," then the market they face will not just be a software market, but the entire labor market.
Within such a narrative, KingSware, relying on years of industrial accumulation, has grown into a leading player in this track.
The founder, Liao Wanli, graduated from the Computer and Application major at Hangzhou Institute of Electronic Engineering in 1993 and immediately joined the technology department of the Zhuhai Branch of the Agricultural Bank of China, working as a core technical personnel for nearly eight years. After leaving the Agricultural Bank of China, he held key positions in financial IT companies such as Yuxin Yichu, Mingtai Information, Yuxin Hengsheng, and Jinzheng Technology.
In March 2016, he founded the current KingSware in Zhuhai.
Originating from banking core systems is a key characteristic of KingSware.
In the 2019 global bidding by China Construction Bank, KingSware competed with international peers such as UiPath, Blue Prism, and Automation Anywhere, and ultimately won the bid with K-RPA in first place. Subsequently, the Industrial and Commercial Bank of China (ICBC), the Export-Import Bank of China, multiple provincial branches of the Bank of China, and the Bank of Communications Credit Card Center, among others, successively became its clients.
As of the end of June this year, KingSware has provided services to over 270 banks, with revenue from the financial services industry consistently accounting for about 70% of its total revenue.
However, on the flip side of its glamorous track record lies continuous financial pressure.
In the first half of 2026, KingSware's revenue reached CNY 65.13 million, a year-on-year increase of 41.7%. However, looking at a longer time frame, the company's growth trajectory over the past three years has been almost flat. From 2023 to 2025, the company's revenues were CNY 217 million, CNY243 million, and CNY 256 million, respectively, with a two-year CAGR of only about 8.7%.
In contrast, during the same period, the CAGR of China's enterprise-level AI solutions market reached 37.1%—enterprises invested more money in AI products other than digital employees.
Its profit performance falls far short of its revenue performance. From 2023 to 2025, KingSware's net loss expanded year by year from CNY 62.57 million to CNY 346 million, accumulating a total loss of CNY 530 million over the three years. In the first half of 2026, it continued to record a loss of CNY 109 million.
Even after excluding factors such as changes in the fair value of financial instruments like redeemable preferred shares, its adjusted net loss re-expanded from CNY 2.95 million in 2024 to CNY 27.13 million in 2025, and further expanded to CNY 56.82 million in the first half of 2026.
Immense Pressure
To understand why KingSware experiences revenue growth without profit growth, it is necessary to first clearly understand its revenue structure.
In 2025, KingSware's project-based revenue was CNY 197 million, accounting for 76.9% of total revenue, of which customized development and deployment services accounted for nearly half of the total revenue, reaching CNY 127 million.
In addition, its subscription-based revenue decreased from CNY 58.73 million in 2024 to CNY 43 million in 2025, with the proportion dropping to 16.8%. The gross profit margin for project-based business was 44.2%, while that for subscription-based business was 75%—the former being less than 60% of the latter.
This is a critical structural issue.
In the context of this company, project-based business is more about "case-by-case" customized services. Under this model, each client requires the implementation team to be on-site for demand research, process configuration, scenario debugging, and on-site acceptance. Revenue expansion is highly dependent on personnel expansion, making it extremely difficult to dilute marginal costs.
Data corroborates this. In 2025, KingSware's outsourced deployment costs increased by 139% to CNY 28.58 million. In the first half of 2026, out of KingSware's 905 employees, delivery and operation & maintenance personnel reached 373, surpassing the 342 R&D personnel.
Such a business model will significantly increase cost expenditures.
First, for every new client expanded, the company needs to match corresponding implementation and O&M manpower. The costs of these personnel are rigid and difficult to dilute with scale growth.
Second, the project-based model also leads to increased pressure on the expense side. In the first half of 2026, sales and distribution expenses increased from approximately CNY 29.21 million in the same period last year to about CNY 59.96 million, a surge of 105.3%, far exceeding the revenue growth rate of 41.7% in the same period. The proportion of sales expenses to revenue surged from around 29% from 2023 to 2025 to 92.1%.
To maintain the effective operation of projects, R&D expenditures also need to be scaled up synchronously. From 2023 to 2025, KingSware's R&D expenditures increased from CNY 55.38 million to CNY 78.77 million. In the first half of 2026, it continued to grow by 34.5% to CNY 42.53 million, accounting for as high as 65.3% of revenue.
The low gross profit margin and high expenditures of the main business mean that pressure will inevitably be reflected on the balance sheet.
As of the end of June 2026, KingSware's cash on hand was only about CNY 72 million, while liabilities related to redeemable preferred shares reached as high as CNY 2.126 billion. These liabilities stem from the valuation adjustment mechanism (VAM) agreements signed with investors during financing—if the IPO fails to materialize, investors have the right to demand the company to redeem their shares.
This means that if the IPO is hindered again, the company will face greater liquidity pressure. Going public has become an inevitable path for KingSware to sustain its operations.
However, even if the listing is ultimately completed, the crux of KingSware's development will remain difficult to resolve in the short term, which is determined by the current development stage of the digital employee track.
Testing Ground
Although many enterprises claim that digital employees "save manpower" or "improve efficiency," most of these effects lack unified standards, and few enterprises can break them down into quantifiable profit contributions in their financial reports.
For instance, in financial scenarios, there are numerous expressions for the value standards of digital employees:
The Industrial and Commercial Bank of China (ICBC) and CITIC Bank adopted the measurement unit of "man-years." The former believes that "the annual workload undertaken by AI digital employees is equivalent to 55,000 man-years," while the latter stated that intelligent models have improved efficiency by over 17,000 man-years;
Bank of Jiangsu used the concept of "work hours." Its 2026 semi-annual report disclosed that various AI-related applications across the bank have cumulatively replaced 1.2 million work hours;
China Merchants Bank (CMB) claimed in its annual report that "AI has saved over 15.56 million hours of manual efficiency."
Regardless of the expression, ultimately, none of the major banks have consolidated the cost-reduction results into actual monetary figures.
The lack of a sufficiently unified evaluation framework means that corporate procurement decisions are more "driven by the wave of the times." The primary motivation is not to fall behind in the industry's AI race, and it has not yet developed to the stage of real efficiency validation and value-driven decision-making.
This procurement model also poses many problems for suppliers. Besides facing greater bargaining pressure, once the product does not meet expectations or fails to generate commercial effects, enterprises easily exit usage and remain on the sidelines, waiting for upgrades.
KingSware's customer retention rate dropped from 74% in 2023 to 66% in 2025, and further declined to 60% in the first half of this year. The average customer unit price fell all the way from CNY 380,000 to CNY 150,000.
Putting these indicators together paints a picture: AI projects have gone live, but clients have not seen sufficient returns, so they choose not to renew contracts or demand suppliers to lower their quotes.
Moreover, this is not the dilemma of KingSware alone.
A global technology report released by KPMG in June 2026 shows that 88% of surveyed enterprises have begun integrating Agentic AI into their systems, but only 24% of enterprises can achieve return on investment across multiple AI use cases.
The first-half 2026 CIO report released by Gartner shows that 83% of CEOs are increasing AI investments, while 59% of AI projects fail to enter the production environment. Among the 11,000 CIOs surveyed, 71% stated that it is difficult to filter out AI application scenarios that can bring quantifiable results.
In other words, everyone thinks digital employees are a good thing, but few enterprises have actually saved money using them, and some cannot even calculate clearly how much money they have saved.
The track appears to have infinite potential, but in reality, almost all participants are moving forward by feeling their way. In such an environment, it is impossible for clients to entrust all their digital employees to a single powerful supplier, making the industry relatively fragmented.
Data from Frost & Sullivan shows that although KingSware ranked first in the AI digital employee solutions market in 2025, its market share was only 3.3%, while the second and third places were 2.9% and 2.1%, respectively, with the top five vendors holding a combined share of about 11.3%.
It can be seen that this is a highly fragmented early-stage market with low entry barriers and far from forming dominant-level barriers.
If the intelligence level of the entire industry undergoes a significant upgrade in the future, a large number of institutions, including financial institutions, will "act according to circumstances" again. Before this, institutions will maintain a relatively indifferent attitude towards suppliers' products. Rather than purchasing digital employee products, it is more like making a "strategic allocation" for the future that must be done but is relatively cautious.
For suppliers like KingSware, this is a period of strategic opportunity in the early stage of industry development, but it is also a continuous ordeal.
Conclusion
The difficulties faced by a company are, in essence, a typical sample of how various tracks are gradually moving from the carnival of concepts to the real validation process of specific business activities in an approaching AI era.
AI has become a strategic direction with strong future certainty across various industries, but tactical approaches are generally far from mature.
Explorers can attract capital attention, enter the capital market, and even secure a large number of client orders. However, even so, everyone is in the early stage of industrial development, with digital employees and KingSware being typical representatives.
For capital and clients, the various AI products emerging one after another at this stage share similar commonalities. They are future-oriented and will inevitably have sufficient development opportunities, but time will always be the key yardstick for measuring all development and benefits.