Vision on AI Impact in Software Development — Part 1: Bert Jan Schrijver
OpenValue’s CTO Bert Jan Schrijver on why developers must embrace AI as a power tool — not a replacement — and what that means for the craft of building software.
Series introduction: In this blog series “Vision on AI Impact in Software Development”, OpenValue experts share their honest, practical perspective on how AI is reshaping our industry. No hype, no fear — just real insights from people who build software every day. We kick off with Bert Jan Schrijver, CTO at OpenValue, Java Champion, NLJUG Community Leader and Duke’s Choice Award winner. Bert Jan has spent over 15 years at the intersection of software architecture, developer communities, and technology leadership.
The Moment AI Got Real
Three years ago, Bert Jan put a large language model (LLM) to the test. The team at OpenValue took the same coding challenge they use to assess developers and fed it to ChatGPT. The results were unmistakable: AI was getting remarkably good at producing working code.
“That was the moment I started thinking seriously about whether AI would eventually replace developers,” Bert Jan recalls. But the picture has since become much more nuanced.
Fast forward to late 2025, and the evidence is everywhere. Bert Jan has been participating in Advent of Code — an annual programming competition in December — for years. This past edition was the first time he noticed something fundamentally different. “I could throw the puzzle description into a model, get working code back, drop it into my IDE, run it, and get the right answer. As a developer, that same puzzle would cost you fifteen minutes to an hour. The AI does it in seconds.”
The productivity gains are real. But Bert Jan is quick to point out something that gets lost in the excitement: writing code has never been the hard part.
The Real Challenge Was Never Typing Code
“We’ve been moving to higher abstraction levels for decades. Building a REST server used to take 300 lines of code. Now it takes three. The real challenge was always somewhere else.” — Bert Jan Schrijver
Software development, Bert Jan argues, has always been about much more than the act of writing code. It’s about understanding what a client actually needs, translating fuzzy business requirements into concrete architecture decisions, designing for quality attributes like security, scalability, and maintainability — and then validating that everything works as intended.
“The real work sits in getting requirements out of people’s heads, understanding the domain, designing the right architecture, defining non-functional requirements, testing, deploying, and making sure things run at scale. Writing code is a fraction of that.”
This is why Bert Jan sees AI not as a threat to the profession, but as a shift in emphasis. The developer’s role is moving — gradually but unmistakably — from execution toward orchestration and validation. You spend less time writing code and more time setting the right context, giving clear instructions, and critically reviewing what comes back.
“It’s like going from being the one who builds the house to being the architect and building inspector rolled into one. You still need to understand every brick — you just don’t lay them all yourself anymore.”
Three Categories of AI in Software Development
At OpenValue, the practical use of AI in daily work falls into three distinct categories.
General productivity tools. These are the non-developer-specific applications that everyone can benefit from. Think meeting transcriptions with automatic summaries, using ChatGPT to brainstorm solutions, or generating first drafts of documentation. “This saves real time,” Bert Jan notes. “After a meeting, instead of spending thirty minutes writing up notes and action items, the AI does it in seconds. And you can share it with people who weren’t there.”
AI-assisted coding. This is where tools like GitHub Copilot, JetBrains’ AI assistant, and Claude Code come in — AI that lives inside your development environment and helps you write, understand, and refactor code faster. “We’re seeing roughly 10 to 20 percent productivity improvement when these tools are used well,” Bert Jan estimates. He’s careful to add a nuance though: the benefit depends heavily on experience level. Mid-level to early senior developers tend to gain the most. Juniors often can’t properly assess the quality of AI-generated output, while true seniors are already fast enough that the marginal gain is smaller.
AI integration in business applications. The third category is building AI capabilities into the software itself — integrating large language models into Java and Kotlin applications using tools like LangChain4j, Spring AI, and running local models with Ollama. “We’ve built quite a few internal solutions this way. The technology isn’t particularly difficult. The challenge is that most clients don’t yet have clear use cases. The possibilities feel endless — and that’s exactly when organizations get stuck.”
Quality Is Non-Negotiable
“It’s fine to use AI to generate code and be faster. But you are still responsible for the results. That hasn’t changed.” — Bert Jan Schrijver
OpenValue maintains clear policies on AI usage in software development, aligned with both internal guidelines and client requirements. The core principle is straightforward: the developer stays in the driver’s seat.
“You can generate code with AI and deliver it faster — that’s great. But you own that code. You’re responsible for validating that it works functionally, that it meets security standards, that it performs well, that it’s maintainable.”
This is the concept of “human in the loop” — and Bert Jan emphasizes it has become more, not less, critical in an AI-augmented workflow. When you write code yourself, quality control happens continuously and implicitly as you make decisions line by line. When AI generates the code, that built-in checkpoint disappears. Validation must become explicit and deliberate.
“You need to feed the AI your non-functional requirements like a shopping list — here’s what it must comply with. And then you need to verify the output with security scans, penetration tests, performance tests, static code analysis, test coverage measurement. None of this is new. But the emphasis on it is growing.”
There’s also a pragmatic balance to strike. “If something takes you an hour to build yourself, but the AI needs eight hours to generate it and you then spend two hours validating and rewriting — that doesn’t make sense. You need to pick the right tool for the right job.”
The Junior-Senior Paradox
“If AI replaces junior developers today, where will the seniors of 2035 come from?” — Bert Jan Schrijver
One of the most thought-provoking implications Bert Jan raises concerns the developer career pipeline. Traditionally, developers grow from junior to senior through years of writing code, receiving feedback, iterating, and learning from mistakes. But what happens when AI can code at the level of a junior developer?
“You could argue that you need significantly fewer juniors if an AI can do what they do. Today’s seniors have the background of ten or more years of hands-on experience — they can read, evaluate, and validate code effectively. But if AI takes over junior-level work for the next decade, how do we develop the next generation of seniors?”
This isn’t just a theoretical concern. It has direct implications for hiring, training, and team composition. Bert Jan himself has always prioritized learning ability over existing knowledge in interviews. “I’d rather hire someone who knows very little but learns incredibly fast, than someone who knows a lot but doesn’t grow. In a few years, the fast learner will have surpassed the other — and they keep getting better.”
He expects the definition of “senior” itself to evolve. “Being senior will increasingly mean being a systems thinker — someone who can reason about the whole, think about architecture, understand how components interact. And perhaps less about being exceptionally productive at writing code line by line.”
When advising a tech lead managing a team of juniors, Bert Jan suggests a hybrid approach: let juniors brainstorm and research with AI tools first, have them create a plan, then review that plan together before any code is written. “What you absolutely should not do is hand a junior an AI tool and say ’let me know when it’s done.’ They can’t assess the quality, and by the time you discover problems, it’s too late.”
What Clients Are Really Asking
As CTO, Bert Jan spends considerable time in conversations with clients across industries. The questions he hears most frequently reveal an industry still finding its footing.
“Most clients are searching. Which tools should we adopt? What should we allow and what not? My developers are asking for these tools — but which ones do we invest in? These are the governance and compliance questions that come before any technical implementation.”
The speed of AI adoption varies predictably. Startups move fast — they’re unencumbered by legacy systems and heavy regulation. Larger enterprises, particularly in financial services and energy, move more cautiously. “In e-commerce, if AI makes a wrong decision, it costs you some money. In banking, it could mean losing your license. In the energy sector, it could mean cities going dark — or worse.”
Digital sovereignty has become an urgent conversation, particularly in Europe. “I’d estimate 90% of AI models used by European companies right now run on American servers. When you see situations like the International Criminal Court being denied access to mailboxes at the request of a foreign government — that’s a wake-up call.”
Bert Jan’s recommendation is clear: private AI. “AI that runs on your own infrastructure, with your own data, in your own data center. Something that keeps running even if geopolitical relations deteriorate. Whether that’s on your own hardware or with a trusted European cloud provider — that’s the direction many organizations are heading.”
Looking Ahead: What Changes and What Stays
“Software architecture won’t fundamentally change. But instead of telling developers what to do, we may be telling an AI what to do. The thinking stays human.” — Bert Jan Schrijver
Looking two to three years ahead, Bert Jan expects several shifts. Proportionally less code will be written by hand, with automated and AI-driven generation taking a larger share. AI will play a significant role in legacy modernization — analyzing massive codebases against defined standards and suggesting concrete improvement paths. And requirements engineering will increasingly involve AI as an active participant, transcribing stakeholder sessions, generating user stories, and iterating on proposals.
He points to a striking example: a company that gave an agentic AI access to its codebase and a customer feedback system with a budget for gathering input. “For a few hundred euros, they got product insights and prototypes that would have taken a team of people several weeks. That’s where things get really interesting — cost-effective, iterative processes that run largely automated, with a human making the final call.”
Yet some things will remain fundamentally unchanged. Software architecture — the discipline of designing systems that serve business goals, balancing quality attributes, creating boundaries and guidelines — will continue to require deep human judgment. “I expect the application of architectural decisions to shift from telling developers what to do to telling AI what to do. But the thinking about what architecture fits which business problem? That stays human.”
The Bottom Line
Bert Jan’s message to developers is characteristically direct: “AI is changing software development. Adapt now!”
Not to surrender control. Not to blindly trust what comes out. But to understand the capabilities, the limitations, and the risks — and use that knowledge to become more effective, deliver higher-quality software, and ultimately enjoy the work more.
“Think of AI as a colleague who handles the tedious tasks. Refactoring code, updating tickets, doing repetitive replacements across a codebase — let AI handle that. Review the pull request, and you’re done. That frees you up to focus on the interesting, creative problems. And isn’t that why most of us got into this profession in the first place?”
Want to learn more about AI Integration in your work? Join our AI Integration for Java Developers training
Bert Jan Schrijver is CTO at OpenValue, Java Champion, and Duke’s Choice Award winner. He has spoken at 40+ international conferences and leads NLJUG, the Dutch Java User Group.
This is Part 1 of the “Vision on AI Impact in Software Development” series. Coming up next: perspectives from other OpenValue experts.
Want a head start on AI in your development workflow? Check out the OpenValue training portfolio — by developers, for developers.
