How Developers of All Skill Levels Can Best Leverage AI

As AI evolves, developers across the board should use coding assistants wisely. Regardless of their experience, it’s important to remember that AI can’t replace human critical thinking or review completely.

Peter McKee, Head of Developer Relations and Community, Sonar

April 24, 2024

7 Min Read
Young concentrated computer developer writing new code
Dmytro Sidelnikov via Alamy Stock

It’s impossible for anyone within the software industry to ignore the unprecedented growth of AI today. Last year, the global market for AI amounted to nearly $200 billion, and it’s only continuing to grow this year, with experts projecting an annual growth rate of nearly 40% from 2023 to 2030. This technology presents developers with an incredible opportunity to leverage its value by optimizing the volume of code produced to address escalating demands. 

However, there’s still a learning curve when it comes to proper, effective AI application in software development. To truly harness new technology, responsible usage is imperative. The more we incorporate AI into the development process, the more developers need to remember that it’s not a magic bullet -- it comes with stipulations. 

Increased AI adoption has come with its challenges. Studies, like this one done by GitClear, note that AI use in the development process has exacerbated code churn; it’s expected to double by the end of the year. This is often a result of mishandling or misunderstanding AI in the software space. This means poor quality, AI-generated code stands to become a large business pitfall and compound technical debt if not handled properly. 

As AI evolves, developers across the board -- of all skill levels -- should use coding assistants wisely. Regardless of their experience, it’s important to remember that AI can’t replace human critical thinking or review completely. While AI stands to boost efficiency and productivity, it cannot replace developers’ work completely. It’s essential to understand the nuances and pitfalls to ensure more work isn’t being created or introducing business risk. That said, best practices depend on the experience level and foundation of the developer behind the AI. 

Related:IBM Talks Bridging the AI Trust Gap with Developers

AI Helps Educate Junior Developers 

Developers starting their careers are expected to know and use AI as a valuable tool in the coding process. But as they continue learning the ins and outs of the industry, now is the time for them to look to this technology as a tool to enhance their education even further. 

One of the great things about AI is that it lowers the barrier of access to coding and allows those who aren’t tenured pros to pitch in on projects and help build quality software. However, a developer who doesn’t yet have years of experience may not be able to tell where code needs refactoring or properly diagnose issues. Because these developers are still building up their experience, they are at a greater risk than their peers of not catching places where AI coding assistants create low-quality, bad code that may be immediately obvious to those with more expertise. 

Related:JPMorgan Chase’s Sandhya Sridharan Talks Empowering Engineers

Part of the problem is that AI lacks the broader context of code. When it presents a solution, that solution may not actually fit the pattern or practices of the code base being used. Instead, it may make assumptions with the consequence of poorer quality. AI can only reinforce the flaws of the inputs -- while it makes it quicker to write code, improper use means that AI coding assistants sacrifice quality for speed. The good news is that AI can help educate and augment learning for those newer to development. It can help junior developers find a solution, reinforce the fundamentals, and explain why or how a piece of code works.  

That said, junior developers cannot rely on AI as a crutch when getting their work done. Instead, they should continue using it to learn the basics and hone their skills, while understanding that software is only as strong, secure and useful as the code behind it. By employing a “trust, but verify” mindset while using these tools, junior developers can have greater confidence in their code. Per GitHub’s own recommendations, developers at all levels have to take responsibility for code reviews and quality checks to make the most of AI code-generating technology. 

Related:Get Ready for Software 2.0

AI Allows Prioritization for Senior Developers 

Because senior developers have the background understanding of the foundations of quality clean code and know how to catch mistakes or inconsistencies junior developers may miss, these professionals don’t run as much of a risk of relying too much on AI. Their seniority means they can ideally better diagnose an issue or find a problem in the generated code before it continues through the software development lifecycle (SDLC). 

For more seasoned developers, AI can work as a shortcut and a time-saving measure, which is critical as demands upon them increase. For example, senior developers can let AI write low-priority code while they focus on the pieces of a project that are higher-value. They’re able to spend their time, effort and attention on the 20% of a project that’s make-or-break, while AI can take care of the 80% that’s more tedious and low-stakes. 

However, senior developers still need to exercise caution. AI-generated code should be approached with a mindset of requiring informed review, ideally combined with static analysis and unit testing, which help ensure the software is fit for production. 

Experienced developers know that the best way to prevent the kinds of bugs and issues that often appear in code, AI-generated or not, is to take a “clean as you code” approach. This means reviewing and testing code earlier in the process to prevent mistakes from making it through the SDLC to deployment. Developers who use this methodology prevent their teams from needing to fix issues later, which only increases cost and results in project delays and technical debt. 

AI Helps Re-Educate Returning Developers 

Plenty of developers have come back to coding after time away, only to discover (to no surprise) that things have changed. Experience gives them a leg up over their junior colleagues, and AI can be a useful educational tool to re-familiarize them with coding practices. 

Even if those returning to coding feel rusty or don’t have much experience with AI, the good news is their baseline knowledge allows them to read the prompts, ask the right questions and understand context in code. While they may be lacking some of the technical expertise of their senior developer peers who have been in the weeds daily, AI allows them to brush up on anything they may have forgotten and accelerate their return to programming. This re-education through AI helps ensure demand can be fulfilled for higher-quality code that may have evolved since they last engaged in coding. 

Anyone coming back to writing code should want to, and must, adapt these tools. AI technology isn’t going anywhere; learning to use it and becoming comfortable with both its benefits and pitfalls can only put these developers in a position to re-learn the basics and feel up to speed with the rest of their peers. 

All Developers Can Benefit from AI 

Regardless of skill level or years of experience, an entire software development team can benefit from using AI tools in their day-to-day work. These tools will only increase in popularity and adoption as they improve. The more businesses depend on software, the more vital it becomes for developers to take advantage of AI to meet demands and get more done, faster. Although generative AI tools serve as vital resources for alleviating developer burnout and enhancing productivity, it remains essential to have a strong grasp of how to effectively utilize them to minimize potential risks. 

Every developer -- regardless of confidence level and seniority -- needs to remember that proper oversight is necessary when it comes to AI. Unchecked use of AI coding assistance is more likely to result in bad, buggy code, and more work for DevOps teams later as they retroactively fix quality and security issues. These tools are not perfect. Microsoft Research, for example, examined 22 coding assistants and found their performance falters against benchmarks. 

To minimize the disruption and risks associated with AI, while still maximizing productivity and innovation, companies should ensure they have usage policies in place that put an eye toward quality control. By using AI tools in tandem with the appropriate checks and testing, developers of all skill levels can then be confident that their code and software are secure, maintainable, reliable, and accessible. Understanding the shortcuts and pitfalls of AI at every stage of a developer’s career journey helps both workers and companies alike make the best use of their investments and ultimately succeed in producing best-in-class software. 

About the Author

Peter McKee

Head of Developer Relations and Community, Sonar

Peter McKee is the head of developer relations at Sonar. Originally from Pittsburgh, but currently residing in Austin, Texas, Peter built his career developing full-stack applications for over 25 years. He has held multiple roles but enjoys teaching and mentoring the most.

Never Miss a Beat: Get a snapshot of the issues affecting the IT industry straight to your inbox.

You May Also Like


More Insights