Amidst the resurgence of tech stocks on the Hong Kong Stock Exchange, another cutting-edge tech firm is on the cusp of going public.

On December 15th, Zhuhai KingWise Artificial Intelligence Co., Ltd. (hereinafter referred to as 'KingWise') filed its prospectus with the Hong Kong Stock Exchange, with plans to list on the Hong Kong Main Board.

As one of the earliest adopters of RPA (Robotic Process Automation) in China, KingWise is not only the nation's largest provider of AI digital employee solutions but also holds the top market share in the financial sector for three consecutive years. Its client base spans all six major state-owned banks in China and over 90% of securities firms.

From 2022 to 2024, KingWise's revenues reached 203 million yuan, 217 million yuan, and 244 million yuan, respectively.

In 2025, KingWise introduced the Ki-Agent enterprise-level intelligent agent platform, aiming to integrate the process expertise, component capabilities, and industry knowledge amassed from years of RPA projects into its intelligent agent offerings.

Next, let's delve into the strengths of this leading 'AI digital employee' firm.

/ 01 /

Beyond RPA: 'Thinking' Digital Employees

If traditional software tackles the challenge of 'record-keeping,' RPA addresses the issue of 'task execution.'

Within institutions like banks and securities firms, there has long been a plethora of highly repetitive yet strictly executed cross-system operations: downloading reports from System A, pasting them into System B, and then sending emails; or a series of processes centered around approvals, submissions, and verifications. These tasks, while not complex, heavily rely on manual execution, consuming human resources and being prone to errors.

KingWise initially set its sights on precisely these scenarios.

Through its K-RPA platform, it replicates human keyboard and mouse operations for software robots, enabling 'digital employees' to perform tasks reliably 24/7, covering core financial processes such as credit card approvals, regulatory submissions, and data verification.

However, with the advancement of large models, purely 'rule-based' RPA has begun to show its limitations. A key factor behind this is the expansion of AI's value proposition—AI is transitioning from a single passive tool to a 'full-process automation solution.'

Historically, AI solutions were mostly 'point tools': document recognition merely converted images to text, speech recognition solely performed transcription, and process automation only executed predefined scripts. Each tool enhanced efficiency individually, but processes still required human coordination, and AI did not assume responsibility for the overall outcome.

Today's AI, however, transcends mere information acquisition. It begins to comprehend business semantics, determine subsequent steps, coordinate multiple systems for collaborative execution, and ultimately complete an end-to-end business process.

This also implies that intelligent agents no longer rely entirely on pre-written rules. They can interpret natural language instructions, break down tasks based on objectives, invoke tools, and dynamically adjust paths during execution.

The resulting transformation is that intelligent agents are gradually replacing scattered 'point automation' to become the unified entry point for understanding intentions, allocating resources, and driving process completion within enterprises.

It is against this backdrop that KingWise launched the Ki-Agent enterprise-level intelligent agent platform this year.

From the early K-APA to today's Ki-Agent, KingWise has primarily focused on three key areas:

The first step is to refine 'execution' into specific modules.

Leveraging years of RPA project experience, KingWise has dissected numerous common business actions into reusable execution units. As of June 2025, the company has accumulated over 4,600 automation functions and more than 1,000 robotic products, covering office software, databases, and various industry systems, essentially encompassing most typical work scenarios.

This means that, at the execution level, agents do not need to 'improvise' but can directly utilize mature, stable, and verified capability modules.

The second step is to 'demystify' AI capabilities rather than encapsulate them as a black box.

KingWise did not merely integrate large models into processes but embedded AI capabilities as components within the platform. As of June 2025, the company has developed 102 AI components, covering areas such as NLP, CV, and OCR, capable of processing unstructured data like contracts, invoices, and identification documents.

These components equip intelligent agents with environmental awareness, goal perception, and dynamic decision-making capabilities. Combined with the deterministic execution of RPA, they form a 'decision-making—execution' collaborative system: models handle judgment and planning, while RPA ensures precise execution.

The third step is to enable intelligent agents to truly 'grasp the business.'

To achieve this, KingWise conducted industry-specific training for large models: fine-tuning them based on open-source models and injecting proprietary industry knowledge bases. As of now, this knowledge base contains over 10,000 structured scenario rules and more than 1,000 industry components, covering highly specialized contexts such as finance and government affairs, effectively reducing model hallucination rates.

More critically, this knowledge is not abstract text but derived from the process accumulation of years of RPA projects. It clearly describes scenario names, operational steps, involved systems, and target outcomes, providing intelligent agents with a stable 'process baseline.'

This means that agents do not reason freely in a vacuum but plan and optimize within verified process spaces. Thus, KingWise has completed its transformation from a financial RPA leader to an 'enterprise-level AI Agent platform.'

/ 02 /

Annual Revenue of 244 Million Yuan, with Financial Sector Contributing Nearly 80%

While the narrative of intelligent agents is captivating, their value to KingWise has yet to fully materialize at the commercialization level.

From 2022 to 2024, KingWise's revenues reached 203 million yuan, 217 million yuan, and 244 million yuan, respectively, with a compound annual growth rate of only 9.5%. In the first half of 2025, revenue declined from 55.48 million yuan in the same period last year to 45.98 million yuan. The company explained in its prospectus that the acceptance pace of some recurring projects was delayed.

This is not a sporadic fluctuation but closely tied to its business model structure.

KingWise's current revenue primarily stems from two models: project-based and subscription-based. Project-based models cater more to customers with high customization needs and complex processes; subscription-based models mainly serve customers with stringent requirements for data security and long-term stability.

Structurally, the company still heavily relies on project-based models. As of 2024, project-based revenue accounted for 69.7%; although subscription-based revenue increased from 20.4% to 24.1% over the past two years, it has not yet become the dominant revenue stream.

However, the rise in subscription-based revenue has indeed improved the company's financial health. With increased standardization, the company's gross margin rose from 42.1% in 2022 to 53.4% in 2024; during the same period, adjusted net loss narrowed significantly from 29.035 million yuan to 2.953 million yuan.

But on the flip side of growth is the continuous decline in average customer revenue.

From 2022 to 2024, KingWise's number of paying customers increased from 384 to 686, but the average revenue per customer dropped from 528,600 yuan to 355,700 yuan.

In contrast, its customer acquisition cost has also decreased. The company disclosed in its prospectus that its customer acquisition cost fell from 291,300 yuan in 2022 to 225,200 yuan in 2024.

Looking back, the company's relatively slow growth is closely tied to its unique customer structure.

As of now, approximately 80% of the company's revenue comes from the financial services industry. The prospectus shows that the company has served over 240 banks (including all six major state-owned banks), more than 130 securities firms, and over 170 other financial institutions.

This achievement is no accident.

KingWise originally spun off from A-share financial IT company Kingdom. In its early stages, Kingdom once held a 51% stake in KingWise, which, after multiple rounds of financing dilution, still stands at 12.14% as of now.

Kingdom itself is one of the highest market share holders in the securities IT sector in China, serving 'large asset management' institutions such as securities firms, mutual funds, and bank wealth management for an extended period. This relationship opened the core customer gateway for KingWise in the financial industry and shaped its earliest and most robust competitive advantage.

However, this same advantage is also restricting the company's growth boundaries.

The financial industry is characterized by long project cycles, high customization, and conservative procurement rhythms, making it difficult to sustain explosive growth. Whether KingWise can replicate its success in non-financial sectors such as government, manufacturing, and energy to build a second growth curve will become a critical variable for its future.

At least from the current data, the upgrade to an 'enterprise-level intelligent agent' has not immediately translated into faster revenue growth.

Nevertheless, as the company emphasized in its prospectus, AI digital employees are still in a high-growth, competitive landscape not yet solidified phase.

Whether KingWise can truly transition from a 'financial RPA leader' to an 'enterprise-level AI Agent platform' ultimately depends not only on the sophistication of its technological path but also on its ability to continuously prove the self-sustaining capabilities of its business model amid fierce competition.

The outcome of this transformation will require time to unfold.

Text/Lin Bai