AI can now generate software in seconds. So why are the companies building these tools still hiring developers?
OpenAI is expanding Codex. AI coding startup Cognition recently raised $1 billion at a reported $25 billion valuation. New AI-powered development tools are emerging almost weekly, fuelling predictions that software engineering may be one of the first professions disrupted by artificial intelligence. Yet despite those predictions, many of the companies building these tools continue to hire developers at scale.
At first glance, the argument appears straightforward. If AI can generate code, and software engineers write code, then it follows that software engineers may eventually become obsolete. The assumption behind many of these predictions is that software development is fundamentally a coding activity, but it’s not.
The value of software development has never been code itself. It has always been the ability to solve problems, with code serving as the mechanism through which those solutions are delivered.
Code is an output. Software is an outcome.
Much of the excitement surrounding AI development tools centers on their ability to generate code. What once required hours of manual effort can now be accomplished in minutes. Functional applications, user interfaces, and workflows can be generated through natural language prompts, lowering the barrier to software creation. But in the rush to celebrate code generation, the industry risks confusing a capability with an outcome. Generating code and building software are not the same thing.
Consider the challenge of building a digital banking application. Generating the code is only one step in the process. Determining how customer data should be secured, how transactions should be validated, how the system should scale under demand, and how regulatory requirements should be met are equally important decisions. None of these challenges disappear simply because code can be generated automatically.
This is because software development has never been defined by the act of coding alone. Code is how those decisions are expressed, not where they originate.
In many ways, the industry is focusing on the wrong bottleneck. Most organisations do not struggle because AI cannot generate enough code. They struggle because AI lacks the context, business logic, and organisational understanding required to make good decisions.
Great Software Begins With Understanding Problems
Before a single line of code is written, someone must determine what challenge needs to be solved, who it affects, and what success looks like. The hardest part of software development is understanding the problem well enough to build the right solution, and this is where humans become indispensable.
Businesses rarely struggle because they cannot generate software. They struggle because priorities compete, customer needs evolve, markets shift, and regulations change. Deciding where to focus, which trade-offs to make, and what success looks like requires context that extends far beyond the code itself.
This is where human-judgement becomes crucial. AI can generate software, recommend technical approaches, and accelerate execution, but it cannot fully understand the business context behind every decision or take responsibility for the outcomes that follow. As the barriers to building continue to fall, the ability to identify meaningful problems and make informed decisions about how technology should be applied becomes even more important.
In many ways, the value of software development is not disappearing. It is moving higher up the stack.
AI’s Greatest Contribution Is Removing Friction
Much of the discussion around AI focuses on automation and which jobs or tasks may eventually be replaced. In practice, the more interesting shift is how AI is reducing the friction between an idea and its execution.
Every major technology wave has done this in some form. Cloud computing removed infrastructure barriers, modern frameworks simplified development, and low-code platforms expanded access to software creation. AI is accelerating that trend by making it faster and easier to move from concept to working product.
For software teams, friction appears everywhere: repetitive coding tasks, testing cycles, debugging, documentation, and countless operational activities that consume time without directly creating value. As AI takes on more of that work, developers can spend more time understanding users, solving business problems, and making better technical decisions.
The impact extends beyond productivity. When the cost and complexity of building decrease, teams can experiment more freely, iterate faster, and bring ideas to market with fewer resources. More importantly, more people gain the ability to participate in the creation process.
Viewed through that lens, AI is not diminishing the role of developers. It is increasing their leverage.
The Future Belongs to Human-AI Collaboration
The debate around artificial intelligence is often framed as humans versus machines. In reality, the organisations creating the most value are those that combine human judgement with AI-driven execution.
AI is exceptionally good at accelerating execution, reducing friction, and increasing productivity. Humans remain responsible for context, judgement, creativity, and decision-making. As software development evolves, those strengths become more important, not less.
The question facing the industry is no longer whether AI can write code. It clearly can. The more important question is what people do once code is no longer the primary constraint.
Beyond Code
The software industry is entering a future where generating code may no longer be a specialised skill. As AI becomes faster, more capable, and more accessible, the ability to create software will increasingly be available to anyone with an idea and the tools to execute it.
That does not diminish the role of developers. It changes where value is created.
As AI takes on more of the execution, human contribution moves higher up the value chain: from writing code to designing systems, from building features to solving problems, and from implementation to innovation.
The future of software development will not be defined by how much code AI can generate. It will be defined by what people choose to build with it.
