In the first year of AI industrialization, will the legal department be the first to "taste the crab"?

In 2025, the first year of AI industrialization is on the verge of arrival, only lacking the "east wind" of key scenarios.

From the increasing number of AI applications autonomously connecting to DeepSeek to the rise of Agentic AI, the rapid development of technology and product forms has brought about the era of AI equalization, seemingly pushing the intelligent transformation of all industries to a critical point.

In the legal technology industry, Fadada, a leading domestic provider of electronic signature products and solutions, has given its own answer based on a decade of accumulation - releasing the legal AI agent product iTerms Pro based on its self-developed large legal model.

Digging deep into the essence of AI industrialization, it's not hard to find that the value of AI doesn't lie in showy skills but in enabling technology to penetrate professional scenarios, especially in legal positions with extremely low tolerance for errors. A good legal AI product needs to amplify professional value to achieve the effect of human - machine collaboration.

Therefore, it's not difficult to productize AI, but the challenge lies in making AI applications truly assist legal professionals in their work.

01. Agent "Breaking the Deadlock" under Compliance Challenges

"Digital transformation in the legal department has been implemented among medium - and large - sized customers for a long time, but the vision of human - AI collaboration driving business processes still has a long way to go."

Mei Rong, the product director of Fadada, told 36Kr that currently, more than half of domestic enterprises have completed the digital management of contracts.

However, at the same time, most of them are still stuck in the shallow end of "online process", with a large number of non - standard contracts relying on manual review and operation, resulting in very low digital collaborative value.

Last year, the in - depth implementation of regulations such as the Data Security Law and the Personal Information Protection Law, combined with the increasingly strict global cross - border data flow supervision, have posed challenges for corporate legal departments to shift from passive response to active prevention and control.

However, general large models are obviously not suitable for legal positions with extremely low tolerance for errors. Most legal AI products on the market are still positioned as single - point tools.

In contrast, iTerms Pro can disassemble tasks like a human, covering scenarios such as intelligent contract review and performance monitoring, legal research, in - enterprise compliance Q&A, and dispute case handling. Through combined AI and multi - modal adaptation presets, it can achieve collaborative planning among multiple agents.

The home page interface of Fadada iTerms Pro

In fields such as law and finance where AI can never replace human experience, Agentic AI can also generate digital employees through CoT (Chain of Thought) and CoA (Chain of Action) to achieve human - machine collaboration.

"One of the deepest feelings we had in the early stage of industrialization is that under the influence of AI equalization, enterprise customers are more proactive than ever in co - building with us. For example, they discuss the idea of implementing legal AI in business scenarios based on their in - depth practice of using AI tools, or provide us with business know - how data to train vertical models to improve the accuracy of large models in business applications."

Mei Rong believes that the first step in the implementation of AI industrialization is to enable different roles in the enterprise to deconstruct requirements through AI's reasoning ability, thus generating more possibilities.

02. AI Elevating the Value of Legal Services: Deconstructing "Replacement" with "Symbiosis"

Under the trend of technology popularization brought by open - source large models, verifying whether the direction of AI applications can be commercially viable is not an overnight task.

In 2019, after five years of exploration in the legal technology field, Fadada officially established an AI product laboratory. At that time, the concept of large language models didn't exist yet, but there was already a lot of knowledge and experience in the legal field that could be precipitated into data. Combined with technologies such as machine learning, OCR, and semantic understanding, it was further used for engineering development and commercial implementation.

Through in - depth research on supervised learning and a large amount of data annotation work, Fadada had already made significant improvements in enterprise risk identification and early warning in the scenario of contract data recognition at that time.

In 2021, Fadada launched a contract intelligent review system based on the life cycle of contract drafting, business negotiation, and transaction structure review, gradually exploring the "right direction" for future legal AI applications.

According to the observation and judgment of Li Lin, the vice - president of Fadada at that time, since large customers had already established their own contract template libraries and standard review processes, they would only introduce AI capabilities in some sporadic legal scenarios. The value of intelligent tools was actually not enough to participate in judgment and decision - making.

After the arrival of the GPT era and until 2024, the costs of training, fine - tuning, and reasoning deployment of large models, as well as their professional generation performance, have all undergone qualitative changes. Fadada also realized that the technology could achieve a leap in "understanding ability".

Based on past research and professional data, Fadada began to self - develop a large vertical legal model and immediately engaged in engineering, looking for breakthrough points that could be intelligently improved around the core scenarios of legal professionals.

At that time, there were almost no large legal model products available for enterprise applications on the market, while Fadada had positioned its application as an upper - layer application that legal professionals could use out - of - the - box.

To achieve this goal, Fadada has carried out a large amount of professional training and optimization from data to algorithms. For example, on the premise of co - building data with enterprise customers, distilling the knowledge precipitated by humans into data is more sensitive and difficult to train than data in production and operation.

At the beginning of this year, the explosion of DeepSeek announced the arrival of the era of large reasoning models. Fadada, a technology company on the "one - way street" of vertical AI applications, also felt unprecedented pressure.

However, the anxiety was quickly relieved: for the legal professional field, legal professional data and knowledge graphs mostly exist in non - standard structures in the daily operations of enterprises and the experience judgments of experts and scholars. Therefore, open - source large models can only be regarded as "the shoulders of giants", but they can't really compare with vertical large models.

The contract comparison interface of Fadada iTerms Pro

Li Lin believes that the mission of iTerms Pro is to truly assist the legal community in their work, achieving an efficiency explosion under human - machine symbiosis rather than replacing legal professionals.

He said that overall, the barriers and evolution directions of future legal AI applications are mainly reflected in the following three aspects:

1. The accumulation of high - quality proprietary corpora and the ability to transform expert experience into algorithms, which are the underlying support for intelligent decision - making.

2. The dynamic adaptation ability to multi - jurisdiction compliance supervision in the context of globalization, which requires seizing the window period of enterprise intelligent transformation through technological innovation.

3. Deepening human - machine collaboration in practical forms such as RAG (Retrieval - Augmented Generation), Agentic AI (Intelligent Agents), and legal digital employees, which not only helps new employees grow quickly but also releases the productivity for strategic innovation for management.

Therefore, it is particularly important to be customer - centric, understand real needs, and carry out customized deployment to realize the value of human - machine symbiosis.

03. Reconstructing the "Strategic Value" of the Legal Department

Compliance under human - machine collaboration is the first - level requirement and implementation method of legal agents, while the deeper pursuit is to reconstruct the value of the legal department.

In a stable market environment, the legal system tends to be mature, and enterprises don't feel that "compliance strategy is growth power". Therefore, the digital transformation of the legal department in many enterprises only stays at the shallow level of contract management, and the legal department has little room for innovation.

The main reason is that there are a large number of "non - standard problems" in the legal field, which take up most of the time and energy of legal professionals. Over time, a functional cycle is formed.

The goal of AI is to release this productivity. Taking Fadada's intelligent contract review scenario as an example, compared with the average 20 - minute manual review time, each contract can save 50% of the workload.

The contract review - risk notification interface of Fadada iTerms Pro

After shifting from inefficient retrieval, citation, and proofreading to strategic optimization, the legal department can find the "breakthrough point".

For example, the demand for going global has greatly highlighted the strategic value of the legal department.

In the process of seeking change externally, the requirements of geopolitics and data security and personal information protection regulations in various countries are far more complex and strict than those in the domestic market. Generally, strategies need to be formulated starting from the enterprise's legal department and run through the enterprise's business lines from production to distribution.

Therefore, to obtain the entry ticket for cross - border compliance, the verification of legal feasibility, the openness and circulation of data review processes, and the legal department's leading of risk control strategies from a professional perspective are crucial for an enterprise's stable global expansion.

"In the legal industry, many customers' expectations and ideas are to learn from others' experiences, ensure the integrity of their own mechanisms, and then explore new businesses."

Huang Xiang, the founder and CEO of Fadada, said bluntly that technology implementation is just a means of departure. The real "attractive fruits" in the legal field still lie in realizing "human value" because the essence of technology is human - centered. While pursuing efficiency, Fadada also always pays attention to issues such as whether the ethical guidelines of technology applications can uphold the legal spirit, how the work process of human - machine collaboration can retain professional judgment, and how local innovative practices can be transformed into universal solutions, so as to maximize human value.

Whether it's learning from others or self - accumulation, for AI to be truly used by enterprises, it needs to connect business processes with single - point AI capabilities. Referring to the logic of the MCP protocol, it is necessary to build a systematic autonomous planning and collaborative ability of agents.

This is why Fadada has always attached great importance to the openness of its products. It has not only opened API interfaces for iTerms Pro but also achieved mutual openness when cooperating with OA and ERP systems with high market shares.

The breakthrough point of intelligent transformation focuses on human - centered thinking, that is, the breakthrough of human - machine collaboration. AI cannot replace humans, but it can push human value closer to the "brain" level.

Currently, AI applications are bidding farewell to the era of looking for nails with a hammer. Whether Fadada's efforts can enable Agentic AI to achieve better commercial verification in the legal field is still worth long - term observation.