AI is dominating conversations in boardrooms and technology teams alike. For business leaders, the promise is compelling: faster delivery, lower costs, and fewer development bottlenecks. But software development has seen this story before. So what can business leaders learn from previous technology trends? Boncode CEO Jan Willem Klerkx shares his insights.
A familiar story for software development
Today’s AI excitement mirrors earlier technology waves. In the 1990s, tools like Microsoft Access promised a future where non-specialists could create corporate systems. Later, 4GL development platforms like Powerbuilder made similar claims.
Initially, these tools worked well for small applications and quick wins. But as projects became larger and business-critical, many organizations discovered limitations around scalability, maintenance, and complexity. Systems often had to be rebuilt by experienced engineering teams.
More recently, low-code platforms like OutSystems and Mendix made similar claims of improved productivity. There’s no question that this is excellent technology; it certainly meets the promise of short-term productivity. But to what extent does short-term productivity outweigh long-term business agility?
As Jan Willem Klerkx puts it, “When things get serious, you need deep human expertise combined with excellent technology. You need software engineers, lead architects, and a thorough build system. If there is one thing we should learn from 4GLs and low-code trends, it is that highly skilled software engineers are still very much needed.”
AI solves only part of the problem
Let’s be clear, Jan Willem Klerkx is not anti-AI. Far from it. He sees enormous value in AI tools for improving productivity and handling repetitive coding tasks. The issue is expectation versus reality.
Research suggests software engineers spend only 16% to 32% of their time actively writing code. The rest involves understanding requirements, communicating with stakeholders, analyzing existing systems, testing, and resolving issues.
That raises an important question for executives and CTOs: if coding represents only a fraction of software development work, how much of the overall challenge can AI realistically solve?
Code generation may be faster, but software architecture, business understanding, and long-term planning still depend heavily on human expertise.
When expectations meet reality
Jan Willem recently saw this firsthand with a prospective client planning an AI-led rebuild of a legacy platform. Initial estimates suggested a one-year project with a €1 million budget.
As development progressed, the team realized the challenge was not simply generating code: it was understanding and redesigning the system itself. Timelines expanded to three or four years, with costs increasing several times over.
The project was cancelled.
The lesson was clear: software transformation is often a design challenge rather than a coding challenge. “It is about understanding. It’s about analyzing. Coding comes at the end of all the thinking,” adds Jan Willem.
AI works best as a tool, not a strategy
AI excels at prototyping, accelerating iterations, and reducing repetitive work. However, once ideas mature into serious products, stronger engineering disciplines become essential. But large organizations still expect bespoke, human-led thinking behind mission-critical outcomes.
For businesses, healthier expectations around AI use may be the key takeaway here. AI is evolving at remarkable speed and offers genuine value, but sustainable software still requires people, processes, and expertise behind the technology.
Want to discuss the impact of AI on your systems and engineering teams? Talk to a Boncode consultant today
