The question of whether AI will make developers jobless is one of the most common things I’ve heard lately. My short and honest answer is: No, AI will not make developers jobless, but it will fundamentally change the nature of our work and put some developers, especially those who cannot adapt to change, in a difficult position. Based on my 20 years of experience in system architecture, network, and software development, this transformation encompasses much more than just writing code.
Once, in an ERP project for manufacturing, I asked AI to generate boilerplate code for a module. Its initial output looked quite good, quickly creating an API endpoint and models for simple CRUD operations. However, I found that AI was completely inadequate at understanding the critical part of the job: the complex business workflow that checks whether a product has passed its final quality control before shipment and affects stock movements between different warehouses, and then translating that into code. AI can form coherent sentences, but it cannot understand the spirit of the business.
How Much Code Can AI Write and What Are Its Limits?
AI models, especially large language models (LLMs), have made incredible progress in tasks such as writing code, refactoring, debugging, and creating test scenarios. Quickly spinning up an HTTP server, generating ORM models from a database schema, or writing a function according to a specific algorithm is now possible in seconds. I’ve greatly benefited from AI when creating simple endpoints for FastAPI in the backend of my side product or designing basic UI components for an Android spam app.
However, there are clear limits to AI’s capabilities. Current AIs struggle to understand complex business logic, organizational flows, or subtle dependencies between different systems. As I’ve seen in an internal banking platform, it’s currently impossible for AI alone to correctly interpret and code layered rules such as security requirements, regulatory restrictions, and approval processes from different departments for a financial transaction. AI only processes the patterns and existing data it’s given; it cannot grasp the deep cause-and-effect relationships or the strategic purpose behind those patterns.
Where Has AI Fit Into My Experiences?
I see AI as a “co-pilot” in my work, never an “autopilot.” Especially when developing an ERP for a manufacturing company, we used AI for production planning optimization. We received predictions from AI to analyze data from operator screens and optimally adjust inventory and shipment dates. This provided an efficiency far beyond manual planning.
However, the job of providing AI with the correct data, training models, interpreting its outputs, and seamlessly integrating these outputs into the existing ERP flow was entirely up to me and my team. In critical system administration tasks like determining index strategies in a PostgreSQL database, setting Redis’s OOM eviction policy, or managing journald limits in the system, the simple command suggestions offered by AI were often insufficient. That’s where real experience and the answer to the “why” question come into play.
How Is the Developer’s Role Changing in the Future?
With the rise of AI, the role of developers will shift towards higher-level thinking, problem-solving, and system architecture. Rather than writing raw code, asking AI the right questions (prompt engineering), critically evaluating AI’s output, and integrating this output into existing complex systems will become more valuable. I can get ideas from AI when designing a ZTNA architecture or examining a BGP routing decision, but I make the final decision based on my own experience and risk analysis.
Especially when working on AI application architectures like RAG (Retrieval-Augmented Generation) or agent patterns, I’ve seen how critical it is to ensure that AI doesn’t just rely on what it has learned, but can also access the specific and up-to-date information I provide. This strengthens the developer’s role as an “information manager” and “system integrator.” The developer of the future will need to know the business domain very well, in addition to technical knowledge; because it will be that domain knowledge that tells AI what to optimize.
Is the Real Risk of AI Joblessness, or Something Else?
In my opinion, the real risk of AI for developers is not joblessness, but resistance to change and loss of competence. If a developer sees AI only as a “code-writing machine” and doesn’t bother to learn and use it, or blindly trusts the code produced by AI, then they can truly lose their competitive edge. Last month, I got a timer config for a systemd unit from AI; it looked correct at first, but it entered an unexpected OOM-killed loop due to the Restart policy. To understand this, I again had to use my knowledge of Linux services and cgroup limits.
So, the issue is not how well AI writes code, but how well we understand, manage, and work with AI. The human brain still possesses adaptability, abstract thinking power, and ethical judgment that AI does not. If we develop these abilities and use AI as an extension of ourselves, we can achieve much greater success in software development.