It's 2026. We have AI tools that generate user stories from a meeting transcript in seconds. We have co-pilots that write code.
Requirements Engineering should be a solved problem by now, right?
Well, not quite.
The tools have gotten better. Much better. But the core challenge remains the same – especially when we look at small and medium-sized enterprises.
AI Is Your Turbo, Not Your Driver
Yes, we use AI. Every day.
Requirements Engineering used to look like this: 2-hour workshop, then 3 hours of sorting notes, structuring them, writing tickets.
Today: Feed in the transcript, "Summarize the key points for the ticket system," done in 5 minutes.
That's brilliant. It takes the grunt work off our hands. We can extract documentation from legacy code much faster or generate first drafts of process diagrams.
But: AI doesn't truly understand the context.
It doesn't know that Mr. Miller in the warehouse has been using a special workaround for 20 years because the old ERP system crashes on special characters. It only sees the "official" process.
AI helps you document faster and find gaps ("There's a missing condition for the error case"). But it doesn't replace the conversation. It just makes it more efficient, because you can focus on the content instead of the typing.
The Knowledge Monopoly Problem (SME Special)
In small and medium-sized enterprises (SMEs), we often encounter a phenomenon that even the best AI can't solve: The knowledge monopoly.
There's that one employee (let's call him Klaus) who knows everything.
- How are special prices for key accounts calculated? Ask Klaus.
- What happens with a return from abroad without a delivery note? Klaus knows.
- Why do we do it this way? "Because Klaus has always done it this way."
When you're building software for a company like this, your job isn't just to write requirements. Your job is to make this distributed knowledge accessible.
That's often painful. For Klaus, because he has to let go and share his exclusive knowledge. And for you, because that knowledge is often unstructured, contradictory, and written down nowhere.
This is where projects often fail: Not because of the technology, but because nobody dares to question Klaus's processes or truly digitize them. No AI can untangle this human knot. You have to do it yourself – with empathy and patience.
Technology Changes, Communication Stays
Whether we use Waterfall, Agile, Scrum, or "AI-Driven Development": In the end, people build software for people.
The tools of 2026 help us build prototypes faster ("Here, click through this – is that what you meant?"). That's a huge advantage over the past, where we spent weeks discussing concepts in the abstract. We can show things instead of just describing them.
But the understanding of the why – why are we doing this at all? What problem are we solving? – that still has to emerge between people.
No AI will tell you: "Actually, we don't need this feature at all if we just reorganize the process in the warehouse."
That was our little series on Requirements Engineering.
From the vertical slice (minimizing technical risk), through the iterative cycle (learning instead of guessing), to the reality in SMEs with knowledge monopolies.
We hope you picked up a few ideas for your next project.
Need Support?
Knowledge monopolies, unclear processes, or simply no time for Requirements Engineering? We help you bring structure – from the first analysis to a working system. Get in touch.