A new set of generative AI tools is transforming not only how we develop software but also who can participate in the process. Some have called this AI-supported development ‘vibe coding’ – the process of using natural-language prompts to create functional code, where people with only minimal technical skills are able to create serviceable, basic applications.
The practice can, however, can lead to poorer quality code. In such cases, AI provides the ‘vibe’ – the suggested structure – and some developers might be tempted to accept all of this without proper scrutiny or consideration.
But conventional development relies on understanding specific programming languages and syntax rules. While vibe coding significantly lowers the barrier to entry by reducing the need to understand every language and development pattern’s nuances, it does not eliminate that requirement.
This tension between making development more accessible and maintaining high-quality code has become a seriously contentious issue in modern software development.
AI is transforming the core meaning of development. It can empower team members to focus on desired outcomes, instead of the deeply granular specifics of this or that implementation.
Recent research, conducted by Stack Overflow, shows that over half of UK developers are using AI for coding. But logic, business requirements and user experience are increasingly the important parts of the role, pushing syntax correctness and deep language expertise into a supporting function. Organisations are increasingly seeking professionals who can bridge product vision with technical execution, frequently without the need to write code.
While vibe coding offers the potential to accelerate development and democratise software creation, any vibe coders must be guided with proper governance – to ensure that speed doesn’t come at the expense of quality and maintainability.
Vibe coding in the agentic AI age
AI-supported development is in its early stages, with vibe coding one of the first steps. The next step will be agentic AI, where fleets of bots can be pointed at tasks with minimal oversight.
Vibe coding with GenAI does not necessarily lead to robust, efficient and maintainable applications. Rather, the results often give merely the appearance of a functional app.
Agentic AI will close this gap. Agents are able to interpret instructions such as “build a customer database” and independently manage the complex technical details, transforming quick prototypes into properly engineered outcomes.
If vibe coding is about code generation through natural-language prompts, agentic AI is about creating an autonomous development ecosystem.
This difference is essential. Vibe coding might help augment an individual developer, but agentic AI comprises a whole AI system that can take on a more proactive, planning and autonomous role in software development, all driven by a specific goal.
Agentic AI systems integrate deeply into developer workflows and can conduct advanced code reviews. They can suggest infrastructure optimisations and adapt to other shifting requirements. Research from Deloitte indicates that one out of every four companies using generative AI will use agentic AI pilots in 2025 and this number could double by 2027.
But bringing the benefits of vibe coding and agentic AI together requires planning.
Organisations must establish thorough security protocols, ensure adherence to data regulations and create clear communication channels between AI systems and current tools.
Despite these challenges, the combined power of vibe coding and agentic AI offers huge benefits for development speed and code quality.
The shifting developer landscape
As vibe coding and agentic AI take on more routine development tasks, the role of engineering is changing.
Less-experienced developers face a steeper learning curve with fewer straightforward tasks available for foundational skill development. At the same time, senior engineers must adapt as AI takes over traditional oversight responsibilities.
The industry is seeing a growing demand for new, specialised workers, such as prompt engineers, who can effectively guide and tailor AI outputs. The most valuable skills now include architectural design, strategic thinking and the ability to work with AI systems effectively.
While these changes may create downward pressure on some roles and salaries, they also create opportunities for developers who welcome AI as a partner rather than a threat. The most successful engineers will be those who can use AI to automate routine tasks, allowing them to focus on innovation and strategic problem-solving.
Organisations that embrace vibe coding and agentic AI will gain a competitive advantage through accelerated development cycles, improved code quality and more logical resource allocation.
Conversely, those unwilling to adapt will find themselves at a significant disadvantage in an increasingly AI-powered development landscape.
Emilio Salvador is the vice-president for strategy and developer relations at GitLab
How to implement vibe coding and agentic AI
Introduce developers to AI tools that increase productivity for routine tasks. Focus on building familiarity, comfort and confidence with AI assistance for coding, documentation and straightforward problem-solving.
Move past just code writing to integrate AI tools into testing, debugging, code review and documentation. Find repetitive, time-intensive workflows where AI can create immediate value with minimal disruption.
Establish clear policies for AI tool usage, including data access permissions, security protocols and quality standards. Outline protocols for how AI systems will share information and collaborate across platforms.
Implement autonomous AI agents for self-contained development tasks. These agents can interpret high-level instructions like “optimise this database query” and independently execute the implementation details while maintaining code quality.
Expand the scope of tasks completed by agents and introduce multiple agents working in tandem on complex projects. Integrate agents across the software development lifecycle and reconfigure team structures to foster cross-functional groups that combine technical expertise with domain knowledge.
Implement systems to track agent performance with clear metrics and correction protocols. Invest in organisation-wide AI literacy through training programs that focus on prompt engineering, AI teamwork techniques and effective system oversight.
A new set of generative AI tools is transforming not only how we develop software but also who can participate in the process. Some have called this AI-supported development ‘vibe coding’ – the process of using natural-language prompts to create functional code, where people with only minimal technical skills are able to create serviceable, basic applications.
The practice can, however, can lead to poorer quality code. In such cases, AI provides the ‘vibe’ – the suggested structure – and some developers might be tempted to accept all of this without proper scrutiny or consideration.
But conventional development relies on understanding specific programming languages and syntax rules. While vibe coding significantly lowers the barrier to entry by reducing the need to understand every language and development pattern's nuances, it does not eliminate that requirement.
