Summary of Key Points
FDE (Frontline Deployment Engineers) have suddenly become popular in the wave of AI, essentially addressing the "last mile" issue in the implementation of AI in enterprises—ensuring that AI models/agents are truly integrated into business operations and can operate continuously. Major companies like OpenAI and Tencent are sending their own FDEs to work directly with clients; service providers such as Minglue Technology are transforming their traditional industry expertise into agent capabilities; and companies like Mengniu are training their own internal "AI pioneers" (in-house FDEs). FDE is not just a new term; it places more emphasis on the ability to apply on-site knowledge back into internal processes compared to traditional software implementation. However, it also faces challenges related to efficiency (leverage) and the dynamic nature of customer needs, ultimately leading to a new division of labor in the enterprise AI market during the Agent era.
What is FDE, and how does it differ from traditional software implementation?
Many people think of FDE as a fancy term that doesn't fundamentally change anything, but it is actually different from traditional on-site engineers:
- Traditional implementation: Similar to an air-conditioning installer—installing and debugging a fixed software solution (such as ERP), then leaving without further involvement with the software development team or subsequent business changes.
- FDE: More like a "custom furniture designer, installer, and after-sales consultant"—not only needing to understand customer needs (e.g., how a bank uses AI for loan approval) but also integrating models/agents into the client's systems. More importantly, FDEs need to document the issues encountered on-site and the successful solutions (e.g., specific approval rules for a bank) to improve models/products for other clients. For example, OpenAI's FDEs are responsible for the entire process from understanding the requirements to the final implementation and must document their methods in an "operational manual" for use by other clients.
In short, FDEs act as a bridge between AI technology and business operations, requiring a deep understanding of both technology and business processes, as well as the ability to turn on-site experience into reusable assets.
Why are major companies and service providers competing for FDE roles?
The main reason is that implementing AI is extremely challenging for enterprises. Companies know the benefits of AI but don't know how to apply it to their own operations (e.g., how fast-moving consumer goods companies can use AI to optimize supply chains, or how banks can use AI to reduce bad debts). FDEs help solve this problem:
- Major companies (OpenAI, Tencent): Want to promote their AI models/agent platforms within enterprises. Sending FDEs means providing on-site training to help companies use AI and collecting feedback to refine their models (e.g., adjusting models based on customer needs for greater accuracy).
- Service providers (Minglue Technology): Used to earn money through custom development, but AI has made coding more affordable, threatening their traditional business models. By combining FDEs with agents, they can turn their industry expertise into customized agents, maintaining their advantages while improving efficiency. For example, Minglue's acquisition of Pulian Software aims to extend this capability to industries like oil and electricity.
FDEs have become a key factor in the competition for the enterprise AI market; those who can help companies effectively use AI will gain a competitive advantage.
The challenge of FDE efficiency: How to overcome the need for more staff?
Traditional custom software development is labor-intensive; the more complex a project, the more engineers are required, resulting in linear revenue growth and low profit margins (e.g., Pulian Software's revenue increased by 40%, but its net profit margin dropped from 23% to 8%). To be profitable, FDEs need to increase their efficiency (how much work one person can accomplish):
- AI helps FDEs save time: Previously, FDEs might spend two-thirds of their time coding and one-third on communication; now, AI can automate coding, allowing them to focus more on understanding business needs, analyzing requirements, and verifying results.
- Reusing experience: FDEs should document successful solutions (e.g., AI approval processes for a specific industry) so that they can be reused for similar clients without starting from scratch.
If efficiency can be improved, the labor-intensive model can become less resource-intensive, enabling one FDE to serve more clients and thus increase profits.
Enterprises are also developing their own FDE capabilities: In-house pioneers vs. external service providers
AI has lowered the barriers for enterprises to develop their own FDE teams:
- Large companies (Mengniu): Selecting "AI pioneers" from sales, R&D, and other departments who understand the business best. These pioneers must learn to build agents and identify areas where AI can be applied (e.g., optimizing milk supply chains). At a higher level, they are responsible for implementing solutions and monitoring outcomes, essentially acting as in-house FDEs.
- Small and medium-sized companies (Xixianji): Use AI tools to develop their own systems. For example, the general manager of Xixianji spent tens of thousands of yuan on an AI management system using Alibaba's Qoder, whereas external customization would have cost millions. Since they understand the business (e.g., how to feed sheep and calculate costs), they avoided the need for repeated communication with external service providers.
However, there are limitations to in-house FDEs: large companies need to address organizational issues (e.g., setting up roles and incentive mechanisms), and small and medium-sized companies may face challenges with system scaling, testing, and maintenance. Therefore, in the short term, both external service providers and in-house FDEs will coexist—external providers helping companies get started, and in-house teams handling long-term operations.
The new market division of labor behind FDE: Who will dominate the Agent era?
The popularity of FDEs marks the beginning of a new division of labor in the enterprise AI market:
- Model manufacturers (OpenAI, Tencent Cloud): Possessing the underlying technology, they aim to integrate models/agents into enterprise operations through FDEs to control the technological entry points.
- Service providers (Minglue Technology): Relying on industry expertise, they transform their knowledge into agents, acting as "business translators."
- Enterprises themselves: By training their own FDEs, they gain control over their operations and reduce reliance on external parties.
The future competition will not focus on which model is the best but on who can effectively manage the production, maintenance, and reuse of agents—e.g., creating agents that can be applied in multiple scenarios or adapt quickly to business changes. FDEs are at the heart of this competition, and their roles and boundaries are still evolving. One thing is clear: in the AI era, understanding business processes will be more valuable than just knowing how to code.
In conclusion
FDE is not just a simple role; it is a critical link in integrating AI technology into actual business operations. Those who can establish this link effectively will gain a larger share of the enterprise service market in the Agent era.