The fast-paced world of software development can sometimes require quick solutions, and GPT-written code is one of them. The ability to copy and paste AI-generated code snippets is efficient, but treating it like a magic bullet can lead to costly mistakes in the long run.
Without understanding when it’s appropriate to use AI, developers can make buggy apps and even hinder their professional growth. A well-thought-out GPT strategy can help them leverage AI effectively to benefit the projects and themselves.
Note: When I say GPT, I mean all LLMs out there.
It’s not practical to eliminate the use of GPT entirely, but there are two groups of people I feel shouldn’t use it in certain situations: mid-level and senior developers.
For mid-level developers with senior aspirations, they can shoot themselves in the foot if they heavily rely on AI-generated code. While tools like ChatGPT excel at solving well-known problems (e.g., standard algorithms or common bugs), they ultimately regurgitate similar solutions without offering creative solutions.
To reach the level of seniority, mid-level developers need to read documentation and think creatively and critically. They need these skills to be the architects of innovative solutions, and GPT does nothing to enhance them. It can even erode them when engineers fall into a cycle of dependency.
Senior developers should not use GPT to fix complex or unique problems. GPT lacks a lot of context when it comes to understanding the underlying architecture because it doesn’t build, run, or debug the code.
I have often found that if the first attempt fails, AI tools often resort to guesswork when prompted further. The code suggestions might point developers in the right direction after a few prompts, but they rarely deliver a comprehensive fix.
GPT-written code certainly has its place. However, using it requires that you define use cases so your team doesn’t misuse it. Doing this can help avoid a lot of costly mistakes, such as a code base full of messy code.
One of the best use cases for AI during development is brainstorming. GPT code can help generate ideas, offering alternative approaches or suggesting libraries that developers might not have considered. It’s like having a soundboard for raw, half-baked, or wild ideas.
For junior developers, when they use GPT with other resources like YouTube tutorials and Stack Overflow, it can be a lifeline. It’s easy for juniors to get stuck on syntax or basic logic when learning the ropes. AI-generated code helps them move forward, providing working examples to study, iterate, and apply to the underlying architecture. It’s not about bypassing learning but actively growing their skills while reducing their reliance on GPT.
Senior developers can use GPT for repetitive tasks they find tedious, such as writing boilerplate code, configuration files, or unit test skeletons. AI can handle these chores, freeing them to focus on high-value work like architecture and system design.
Similarly, AI can act as a pair-programming partner, offering quick drafts of code for review and refinement, enhancing productivity without sacrificing quality.
On top of well-defined use cases, there are other things you can do to ensure that your AI strategy goes smoothly during development.
Developers need to check GPT-written code constantly, making code reviews necessary. If you’re already using pull requests to ensure the code meets requirements, that’s a good thing. But I strongly recommend letting your developers work without pull requests and switching to code discussions instead. That way, you can avoid unnecessary delays and context switching – two major issues that kill a developer’s flow.
Furthermore, to embed these practices within your company, mentorship is essential. If you don’t have a developer who is experienced with the best AI practices, consider bringing in a professional mentor to bridge the gap between blind reliance and effective use on all levels of the organization.
A mentor can teach juniors how to interpret AI outputs. They can also guide mid-level developers toward independent problem-solving and instill the soft skills needed to be seniors (e.g., leadership, communication, collaboration). Even seniors can be mentored in how to leverage AI to streamline their workflow so they don’t stretch its capabilities.
ChatGPT and similar tools are useful, yes, but they can make a lot of mistakes. When used wisely—for brainstorming, repetitive tasks, or supporting juniors—they enhance productivity and creativity. Setting guidelines and investing in mentorship is key to supporting your teams, ensuring they don’t use AI as a crutch.