I particularly agree with the last sentence of this post: "The tools themselves aren't difficult; the difficulty lies in translating the client's vague needs into concrete processes that the agent can execute." When I was an SAP consultant providing operations and maintenance services to the sales department of PetroChina, when government agencies submitted development requirements, I wasn't actually responsible for writing the code, but my most time-consuming task was: translating the client's business ideas into development documents and business process diagrams that programmers could understand. Clients usually only say one thing at the beginning: I want to implement this function. But if they don't continue to ask: What problem does this scenario solve? What is the ultimate result they want to achieve? What are the criteria for success? What is the specific process? The programmer simply cannot develop it. Therefore, the real difficulty in many technical projects is never the technology itself, but: translating vague needs into executable processes. This is also why in many teams: people who understand the business are often more valuable than those who only understand the tools. And in the AI/Agent era, this ability will only become more important.