“We just need to get better at defining what we want the AI to look for” » Interview with Allard Buijze

Before taking the stage at GoTech World 2025, we sat down with Allard Buijze, Founder & CTO at AxonIQ, to talk about software architecture, AI-assisted coding, and what it takes to build scalable systems in an AI-driven world.

Allard Buijze is Founder & CTO at AxonIQ and a software architect with deep expertise in scalability and high-performance systems. Over the years, he has helped organizations navigate complexity and make future-proof technical decisions. 

With that background in mind, it’s no surprise that his relationship with code started early and with curiosity. “I’ve always loved the creative aspects of programming,” he says. “As a kid, I could imagine things I wanted the computer to do and then all I needed to do was find out how to tell it to do that.” 

That mindset followed him throughout his career. Even when he initially tried to keep programming as a hobby and chose aerospace engineering as a profession, software remained central. 

“I quickly realized that programming was more than a hobby,” Allard explains. “Even when programming professionally, you can still put in the same amount of creativity. I managed to become a professional developer that flies around the globe.” 

Scalability Looks Easy on Paper Until Reality Hits 

One of the hardest lessons Allard learned came from real-world systems, not theory. “Things that work for one user, or just a few, don’t necessarily work for thousands of simultaneous users,” he says. “Scalability is easy to reach on paper, but in practice, it only takes a very small bottleneck or a false assumption to severely limit it.” 

Testing, he adds, is essential and often misunderstood. 

“I’ve seen scalability tests that were actually testing the scalability of the test code, rather than the production code,” Allard recalls. “That’s why measuring and monitoring everything, especially in production, is so important to validate assumptions.”

Why AI Thrives in Some Codebases and Fails in Others 

With AI now writing, reviewing, and debugging code, Allard believes success depends largely on context. “Whether LLMs can thrive depends predominantly on context,” he explains. “An LLM is essentially a big bag of words that predicts what comes next based on patterns it has seen before.” 

In software development, this becomes critical. “If we focus an LLM on a specific task and limit the amount of context it needs, it will often do the right thing,” Allard says. “But if instructions are vague, or the codebase it needs to scan is too vast, chances are it’s not going to do a good job.” 

This is why architecture plays a key role. “We need to rethink how we structure our code,” he adds. “Vertical, functional slices tend to work much better than systems organized purely around technical layers.” 

The Three Capabilities AI-Ready Systems Can’t Live Without 

From Allard’s perspective, systems that want to thrive in an AI-driven world need three essential capabilities.

“For systems to thrive in a world of AI, they need to provide context-rich data, auditability or explainability, and scalability,” he says. Context-rich data allows AI to make better decisions. “AI can consume and correlate vast amounts of information in a very short time,” Allard explains. “The richer the data, the more value we can unlock.”

Auditability becomes critical as autonomy increases. 

“If AI makes decisions, we need to know exactly what it did, when, and why,” he says. “And if it ever makes a rogue decision, like deleting your CRM database, you’d better have a way to roll it back.” 

Scalability may be the biggest challenge of all. 

“The rise of AI agents will massively increase system load,” Allard explains. “Some call it the fifth wave of computing. Estimates suggest usage could increase ten to a hundred times, some even say a thousand times. If systems can’t scale, they’ll become obsolete very quickly.” 

Why AI Will Make Architecture More Important, Not Less 

Some engineers fear AI-assisted coding will reduce the relevance of software architecture. Allard strongly disagrees. “I think AI will separate programmers from software engineers,” he says. “While AI takes over more of the coding, deciding what to build and how to structure it at a high level will remain a human responsibility for much longer.” 

Architecture, in his view, becomes the differentiator. 

When AI Maintains More Code Than Humans 

Allard believes AI will surpass humans in code maintenance faster than many expect and he welcomes it. “Personally, I actually hope so,” he says. “Maintaining code, especially your own code written more than six months ago, is terrible.” 

AI’s advantage lies in speed and comprehension. 

“AI can read and analyze code incredibly fast,” Allard explains. “It’s much more efficient at maintenance than we are. We just need to get better at defining what we want the AI to look for.” 

A Final Thought for Engineers 

For developers worried about their future, Allard offers pragmatic advice. 

“Yes, AI will take over parts of your job as you know it today,” he says. “Things will change, but we are needed to change the profession ourselves.” 

His recommendation is simple. “Get acquainted with AI as much as you can. By using it, you’ll understand what it can do and what it can’t. Make sure you can do what it can’t.”