Technology has always played a fundamental role in helping to evolve workflows, including the work of software developers. In the 1950s and 1960s, the role of computer programming was linked to punch cards. These stiff paper cards, with patterns and holes in them, provided a tactile version of code, which when fed into a computer, would perform a pre-designed function. While cutting edge at the time, the debugging process was lengthy, and the system ultimately deemed inefficient.
The pursuit of efficiency has driven over 70 years of further coding advancements, creating the complex and intricate digital systems that we have today. The growing importance of artificial intelligence and the ongoing debate about its role in driving developer efficiency is the latest in these advancements. Recent industry surveys suggest that as many as 76% of developers are using or planning to use AI in their development process, to increase productivity, speed up learning and improve accuracy.
Clearly, software developers have enthusiastically adopted AI. However, the emergence of so-called ‘vibe-coding’, the use of AI prompts to create new applications, has raised several questions. How best can AI be incorporated into existing workflows, how do we provide AI with the best platform for success to drive these productivity gains, and equally, how will this impact the development of engineering skills moving forward?
Using AI as an Intern
While technically not new, enterprises are still figuring out how to make the best use of AI in a safe manner. While it can have beneficial implications for task automation, writing and debugging code, and suggesting architecture, it is not without its flaws. Ultimately, AI systems are only as good as their prompts and the data they use. There is risk that by delegating tasks to AI agents or tools, end-users become less diligent over the end results. This opens the door to code mistakes or phantom actions, which ultimately can have serious operational impacts if left unmanaged.
The rush to adopt AI tools has outpaced the implementation of guardrails or the risks involved from overreliance. The proactive integration of agentic AI needs to be managed as the technology continues to develop, and human oversight remains crucial. In essence, developers looking to drive the best results should be treating the technology as an intern. Ensuring maximum efficiency requires detailed prompts, positive reinforcement and iterative mentoring. AI is a powerful tool, and it’s at its best when enhancing, not replacing, human-led expertise.
Driving better AI outcomes through sandboxes
Beyond human-oversight, sandboxed environments can help to safely integrate AI into current developer workflows. These provide a controlled environment to write code, experiment, and explore different ‘what if’ scenarios. Within these sandboxes, developers can give AI tasks without the risks of cascading errors that could cause costly operational issues.
The potential outcomes for developer teams is huge. Bolder ideas often lead to better products and services. Whereas traditionally these may have taken days or weeks for teams to create, AI can often complete these actions in minutes. Developers will be able to see what works and rule out what doesn’t, freeing them from the risks that come when implementing new ideas. When AI has performed a task within the sandbox, a developer can take the elements that work and implement them into their work-in-progress code.
Getting AI to produce outputs at this level still requires collaboration. Often, this means checking if the prompts it is given are accurate for the task and tweaking it where needed, then having colleagues review the generated code before it is merged.
Soft skills are the future of coding
With AI seemingly able to shoulder the resource intensive aspects of programming, there is debate over the future of software developers and the skillsets that they require. The conversations around ‘vibe coding’ have only exacerbated this. While technical skills will always be needed, both to monitor the outputs of AI agents, and review the code as needed, there is a growing value in developing a wider range of soft skills too.
Empathy, creativity and communication are becoming core fundamentals of a well-rounded developer, opening growing potential for those with humanities and philosophical backgrounds to succeed. Ultimately, placing less focus on manual coding will open the potential to think more about what success looks like, the customer issues that the company is trying to solve and the best prompts and approaches to get there.
Technology in the workplace has come a long way from punch cards, and the evolution of programming practices has taken an exciting leap forward with the proliferation of AI. Driving toward maximum efficiency whilebalancing this with safety, requires a methodical introduction of AI into developer workflows, with human oversight over AI-delivered code and sandboxed environments to test the feasibility of AI outputs. Over the long-term, AI is likely to shoulder much of the manual coding processes, evolving the role of the developer and increasing the importance of soft skills.
