In February, we discussed a thought: in the medium to long term, improving efficiency through technology is the trend, and we need to consider which players will ultimately win? Is pure AI capable of disrupting all industries, or are industries/industry giants that actively embrace AI and have the potential for complete transformation more likely? I have recently discussed with friends that AI will indeed "eat up" a large number of generic SaaS, but the more vertical and in-depth the SaaS within the industry, the greater its value and may even become more difficult to replace, 1. Why is general SaaS dangerous? They often do horizontal tasks (writing code, copywriting, making PPTs, simple customer service), which can be improved purely by improving the original model capabilities, And AI big model manufacturers take top-level models+ready-made connectors (G Drive, Slack, Salesforce, etc.)+simple agentic orchestration layers. This type of work can be replaced, 2. Why is the vertical SaaS moat stronger? Industry professional data and "tribal knowledge" are almost unavailable. The training data of the general big model is the public Internet plus+part of the enterprise license data, but a large number of real industry know how are not written, word of mouth, and unique within the company: 1) Invisible constraints in the supply chain of manufacturing industry, equipment testing and maintenance, etc; 2) Exclusive production modes and special process flows in the chemical industry; 3) The unique underwriting philosophy and exception handling rules of insurance companies 4) The approval process between departments in hospitals and gray areas of medical insurance policies These things are not in any public dataset, and even the strongest general AI cannot learn them. Unless you run the real production process and constantly collect human correction signals, the model will always be "almost". The more vertical it is, the more it requires "system level" rather than "tool level" General SaaS is often a "small tool stacked on existing systems" (writing emails, generating reports, simple agents), while vertical SaaS is an end-to-end workflow system that spans legacy systems, human approval, compliance auditing, and P&L direct impact. It is also necessary to be familiar with the industry, production processes, and technological processes unique to the enterprise. In this part of the work, the proportion of deterministic software engineering+governance+data flywheel far exceeds that of the model itself, making it difficult for AI models to fully cover all aspects. Every time a customer is served and a real process is run, an additional "operational memory" is accumulated (which redline was changed by the customer and which exception human intervention occurred). After accumulating 100 customers x 1000 transactions, the accuracy and intelligence of your system in that vertical domain will significantly surpass the capabilities of general models in that vertical domain. From this perspective, vertical SaaS is actually the best "amplifier" and "delivery layer" for AI - AI provides intelligence, vertical SaaS provides industry context, industry proprietary data, specific experience and specifications, compliance, integration, and ultimate responsibility. Companies that use AI as an "engine" and vertical industry processes, data, and responsibilities as a "body" will also be future winners.