Authored by Shane McAllister, Lead Developer Advocate (Global) at MongoDB

Artificial intelligence has the potential to redefine how we work in a way not seen since the Industrial Revolution. In software engineering, this is especially topical as AI-powered coding tools continue to proliferate across developer teams. While these tools save huge amounts of time by automating parts of the coding process, there’s a worry that it could come at a cost. A study by Google found that an over-reliance on these tools is leading to an overall erosion of code quality by increasing duplicated code with minimal refactoring and added complexity. In addition, the tech industry has been debating whether automating coding inhibits younger developers’ ability to understand the underlying theory of the craft, which in turn could impact their ability to think critically and creatively in development environments. 

This is not to say the profession is facing some kind of brain drain. Code development and software engineering are fluid and constantly evolving – remember punch cards and machine code? No, I don’t either, Google it. We’ve come a long way from feeding in paper tape with holes in it, through assemblers, pseudo-code, and finally high-level programming languages to get computers to do what we want. And with each advancement, it’s always touted that the new generation of coders don’t know how good they have it, and are at risk of not understanding first principles and core concepts. AI is simply next. 

However, like all tools, AI coding tools must be used wisely and are only as good as their foundational training and knowledge base. To preserve key skills in logic, problem solving and coding fundamentals, engineering leaders must establish clear guidelines about AI use, actively cultivate cultures of continuous learning and mentorship, and embrace the era of AI-augmented development. After all, those fundamentals don’t change, even if the tools we use do. 

The importance of guidelines 

With AI tools becoming a new normal, business and engineering leaders must learn to work with this technology. AI coding tools must be seen and treated as a tool that allows developers to be creative, think critically, and solve complex problems, instead of becoming a crutch. In practice, this means setting clear policies for where AI can and can’t be used, having an approved list of AI tools, and training developers on how to use AI compliantly to avoid putting sensitive information into a third-party tool. 

A senior engineer working with a small group of junior developers equipped with AI coding tools to develop a new function may find it helpful to implement a “show your work” mandate to ensure automation does not compromise learning. Any significant code snippet, business logic or architectural design generated by AI would need to include an explanation from the developer on why the AI’s suggestion was chosen and how the solution was validated. Similarly, code reviews should specifically scrutinise AI-generated portions. Developers should be ready to answer questions like, “What was the thought process here, beyond the AI’s suggestion?”, “How was this logic verified?” or better still “Is the suggested code as concise as it can be?”. These measures allow developer teams to streamline the coding process, while at the same time preserving their ability to reason and think creatively and critically about their code. 

New tools, new skills 

Many notable inventions have prompted our workplaces – and societies – to reprioritise various skills over time. The rise of applications like Google Maps made navigation easier and a lot more efficient. For some, that meant a major change to how they work. Taxi drivers no longer needed to memorise every road in New York or London, but by no means did it make them redundant. Instead, the ability to quickly build rapport and add that human touch by acting as an unofficial guide became more important. 

In the same way, engineers are likely to see other skills become more important. Instead of getting bogged down in boilerplate code and syntax, developers will need to become experts at breaking down complex, ill-defined problems into logical, solvable parts that can either be addressed through AI or human ingenuity. 

With AI handling more of the low-level (and often mundane) implementation, developers will be freed up to hone their bigger-picture thinking, like designing scalable, robust, and maintainable systems. Luckily, engineers are already lifelong learners, and in many ways, this will simply be another stage in this learning as their day-to-day takes a new shape. 

AI in action 

If applied correctly, AI coding tools can become a force multiplier for what teams can achieve. At their best, they could allow smaller teams to undertake more ambitious projects than their headcount could previously allow. For example, let’s imagine a hypothetical project: building a cloud platform for a city council. For a small team of engineers, AI coding tools provide a head start by quickly setting up the project’s foundations, such as the code, the APIs, and the infrastructure. And instead of it taking weeks, this is exactly the sort of repetitive, pattern-based work that coding tools can automate to free up developer time. 

However, human expertise remains firmly front and center. AI coding tools present the starting points, but it is still crucial that human engineers refine those suggestions for scalability, performance, and ensure it’s easy to maintain in the long run. In other words, even as AI coding tools become great “doers”, the role of humans as the “thinkers” directing the tools remains undisputed. 

It will undoubtedly take active effort to ensure AI coding companions are an evolution of our toolkit, rather than a threat to human ingenuity. But solving complex challenges is at the heart of what engineering is all about, and in this case, clear guidelines and a culture of continuous learning are the surest way to ensure that AI tools sharpen our engineers’ critical and creative abilities – not dull them. 

Photo via Unsplash