Summary of Key Points
This article discusses the challenges in implementing AI in enterprises: while AI significantly improves efficiency at the individual level (for example, in writing copy or researching information), most projects fail to meet expectations when applied at the organizational level. The root cause lies in the fact that the "implicit knowledge" accumulated over decades by companies—such as practical skills and industry insights held by experienced employees—is not digitized, making it unavailable for AI to utilize. Kingsoft Office, however, has successfully transformed this individual expertise into an organizational asset by extracting real-world data from within the company (e.g., legal reviews of emails), thereby solving core issues like contract auditing. The article highlights that the focus of competition in the AI for enterprises sector is shifting from individual use cases to organizational office applications, with a consensus that activating a company's own implicit knowledge is the "last mile" towards successful AI implementation.
Why Doesn't Enterprise AI Work as Well as It Does at the Individual Level?
Many companies find that while employees can quickly generate reports using tools like ChatGPT, the AI projects they invest in often prove ineffective or even counterproductive.
- Data Says It All: Gartner research indicates that 70% of AI projects fail to meet their objectives, and Bain suggests that although investment in AI is increasing, the returns are not keeping up.
- Where Does the Problem Lie? Wu Qingyun, Vice President of Kingsoft Office, explains that when AI is integrated into outdated organizational structures, its limitations are amplified. More importantly, valuable expertise—such as senior lawyers identifying problematic contract clauses or project managers recognizing issues in data—is stored in employees' minds, not documented or digitized. Without access to this knowledge, AI struggles to perform effectively in critical business tasks like contract auditing.
Kingsoft Office's Innovative Approach: Extracting "Golden Knowledge" from Email Logs
Kingsoft Office’s breakthrough in contract auditing did not rely on more advanced models but on leveraging the expertise of its own employees:
- The Challenge: Legal reviews require meticulous page-by-page analysis, often resulting in lengthy overtime. However, the review criteria are not documented in manuals but reside in the minds of experienced lawyers.
- How They Solved It: After observing that lawyers would send emails detailing their findings (e.g., "This clause is a standard practice and does not need to be changed" or "The wording could cause payment delays"), Kingsoft extracted these insights from hundreds of thousands of emails and attachments. These were then transformed into actionable rules using AI, creating a comprehensive knowledge base for contract auditing.
- Collaboration Model: The CFO brought in the most senior contract experts to work alongside engineers. Experts reviewed the AI-generated results and provided immediate feedback, allowing engineers to update the rules in minutes. Previously, IT projects would take three months each for research, development, and modification; now, both teams share the same KPIs, enabling rapid iteration.
- The Outcome: Lawyers no longer need to work overtime regularly, and their initial skepticism has turned into complete acceptance of the AI’s effectiveness. More importantly, the company has transformed the individual expertise of its employees into a tangible asset that can be passed on within the organization.
The Shift in the Enterprise AI Landscape
The focus of competition in the AI industry is no longer on creating smoother interactions but on extracting valuable information from documents:
- The Real Difference: The most valuable knowledge lies not in instant conversations but in frequently revised contracts, reports, and manuals. Those who can effectively extract this information are one step closer to successful AI implementation.
- Kingsoft’s Advantage: As a company specializing in office software, Kingsoft is well-versed in dealing with these documents, giving it a significant advantage in serious applications like contract auditing.
Industry Consensus: Activating Implicit Knowledge Is the Key to Successful AI Implementation
The industry consensus is that the key to successful AI implementation lies in activating the implicit knowledge within enterprises. Microsoft’s Copilot, as well as DingTalk and Lark, are adding AI features for enterprise use.
- The Critical Point: Different companies may adopt different approaches, but they all agree on one thing: the success of AI in enterprises depends on how quickly they can transform the hidden knowledge in their employees into digital rules that can drive business processes.
In One Sentence
For AI to be truly effective in enterprises, focus not just on purchasing computing power or upgrading models. Instead, extract and digitize the valuable expertise held by your employees—this is what will enable AI to provide real value and make a meaningful difference.