Summary in Plain Language
This article explains a fundamental shift in the software industry during the AI era: Over the past 40 years, the golden rule has been “sell the same code millions of times.” This meant investing heavily in developing software, which could then be replicated at almost zero cost and sold to thousands of customers. Companies that offered customization or sent engineers to clients were seen as having poor scalability and unattractive business models. However, with the advent of AI, the cost of writing code has dropped significantly. Instead, the real complexity of business operations—hidden in outdated systems, traditional practices, and unwritten rules—has become the most valuable and scarce resource. Frontline Deployment Engineers (FDEs), who are stationed at clients to understand the actual situation and provide feedback to the headquarters for product iteration, have become highly sought-after. This is not just about creating a new role; it represents a complete shift in the industry’s business paradigm from “copiating software” to “adapting to the real world.”
---
Detailed Explanation
1. The “Unwritten Rules” of the Software Industry Over the Past Four Decades
The key to making money in the software industry was to make customers adapt their processes to the software. The standard for evaluating a product was simple: could the software prevent customers from needing to contact the company’s engineers and allow them to register, configure, and use it on their own? For example, if you run a milk tea shop using a SaaS system that requires setting up discounts one day in advance, the vendor would not modify the software to accommodate a last-minute discount offer. Instead, they would suggest changing the discount time to the previous day, citing it as best practice. This approach was applied from ERP to CRM systems, where 100 companies’ diverse processes were compressed into 3-5 standardized templates, forcing customers to change their work habits to fit the software. This allowed vendors to earn millions from the same code, resulting in high profit margins that attracted investment from the capital market.
On this logic, providing on-site services or custom development was considered “politically incorrect,” as it implied that the product was poorly designed and relied on manual labor, similar to consulting firms, which was not considered sophisticated.
2. The Impact of AI on Costs
With the rise of generative AI, the cost of writing code has dropped to a fraction of what it used to be. However, the complexity of the real world has not decreased; instead, it has become more expensive to address. For instance, developing a procurement system used to cost millions, but now, with AI, the actual cost might only be 10,000 yuan, with the remaining 990,000 yuan spent on understanding specific issues like why financial approvals are not recognized by the system or why certain fields remain unfilled for years. These problems do not disappear with AI; rather, when AI takes over practical operations (such as payment processes), they become critical challenges. The most valuable aspect of software is no longer just developing it but ensuring it is effectively used in the client’s environment.
3. FDEs: More Than Just Implementation Consultants
Many people think FDEs are just traditional implementation engineers who set up software at clients’ locations. However, their role is different. They go to the client’s site to understand all the unwritten rules, historical issues, and power dynamics, translate them into actionable requirements for the software, and provide feedback to improve the product. This is like a customer service agent in a video game: instead of fixing individual issues, they report to the developers to make game mechanics more user-friendly. This process, described by Palantir as “human reverse propagation,” means that AI models are improved based on real-world data, making the software more useful than those developed in labs.
4. The Change in Valuation Logic for the SaaS Industry
In the past, the capital market looked down on software companies’ service revenues, believing that more staff meant lower profit margins, similar to consulting firms. But with AI, the efficiency of FDEs has increased significantly. A project that used to take six months with ten engineers can now be completed in three weeks with one FDE and AI assistance. Sending FDEs to clients is no longer a cost burden but an investment in a valuable asset. By understanding common issues (such as outdated systems and compatibility issues), companies can create standard solutions that can be applied across multiple clients, reducing costs and enhancing scalability.
5. The Competition of the Future
In the future, the competition will not be about who has the largest models or the most code but about who has experienced the most real-world challenges. The most valuable engineers will be those who can quickly identify fake requirements, avoid critical issues, and understand common pain points in unfamiliar industries. The evaluation criteria for companies will shift from focusing on code to their ability to understand and adapt to the real world. While large models, computing power, and code are becoming more scalable, the challenges of the real world—such as system bugs and power dynamics—cannot be automated. Companies that have experienced the most real-world problems and have developed corresponding solutions will have a competitive advantage in the AI-driven industry.