I'm not a software developer.
I'm not a data scientist.
And I'm not someone who understands the latest AI models down to the smallest technical detail either.
I'm a project lead.
Someone who has spent years trying to make processes better, get systems to work together, and make sure information ends up where it's supposed to.
When generative AI broke through, I thought what many people thought: “This is going to completely change how I work.”
And it did.
But probably not in the way most people think.
AI can do a lot. An incredible amount.
Over the past months I've worked intensively with Claude, ChatGPT and other AI systems.
Not to write some nice-sounding text.
But to work through complex problems, build analyses and structure ideas.
What struck me most?
AI processes in a few minutes an amount of information that would take a human days or even weeks.
That's impressive.
But that's exactly where a misunderstanding starts.
Many people think AI is some kind of magic box.
You feed it some documents, ask a question, and automatically get the perfect answer.
My experience is different.
AI needs direction
AI is remarkably good at recognizing patterns.
But it doesn't automatically know which pattern matters for your situation.
That means you constantly need to be able to judge:
- Is this answer actually correct?
- Is context missing?
- Is it using the right assumptions?
- Is it looking at this from the right angle?
And that's where the real challenge sits.
Not in using AI.
But in guiding it.
You don't have to be a developer…
…but you do have to learn to think like someone who brings structure.
I'm not technically strong.
Sometimes I don't know the right terminology.
Sometimes I don't know exactly how a particular technology works.
That sometimes makes working with AI difficult.
Because if you don't know the right concepts yourself, it's hard to steer AI in the right direction.
I often started with a vague idea.
After several iterations I eventually got where I wanted to be.
Not because AI suddenly got smarter.
But because I myself understood better what context was missing.
What information I'd forgotten to include.
What assumptions were wrong.
AI got better because my questions got better.
Domain knowledge still makes the difference
What surprised me most, maybe, is how important domain knowledge remains.
AI has an enormous amount of general knowledge.
But your organization… your processes… your customers… your way of working… AI doesn't know any of that.
You have to bring that.
Only once you add that context does AI actually start creating value.
And even then, human oversight stays necessary.
Because only someone with experience can usually sense, intuitively, when an answer “sounds right” but actually isn't.
The quality of your data determines the quality of your AI
There's an old saying in IT:
Garbage in, garbage out.
It's more relevant today than ever.
When you let AI work with clear, well-structured, up-to-date information, you usually get surprisingly good results.
But work with documents scattered across dozens of locations… versions that contradict each other… unclear naming… or missing context… and AI first has to go hunting for the right information.
And just like a human, it then makes wrong interpretations faster.
Good data isn't a luxury anymore.
It's a precondition for using AI successfully.
AI doesn't replace collaboration
What may have taught me the most is that successful AI projects rarely revolve around one person.
It's about bringing different kinds of expertise together.
- Business knowledge.
- Process knowledge.
- Technical knowledge.
- Data knowledge.
And the skill to keep redirecting AI, again and again.
Not one big prompt.
But dozens of small improvements.
Iteration after iteration.
Until the result is right.
AI makes experts stronger
Sometimes you hear that AI will make everyone an expert.
I believe the opposite, actually.
AI makes experts more productive.
It speeds up thinking.
It helps connect the dots.
It processes enormous amounts of information.
But it doesn't take over the responsibility.
That stays with the human.
And maybe that's the most important lesson I've learned these past months.
Not that AI can do everything.
But that good results still start with a clear goal, quality data, and people with enough knowledge to challenge AI when it's wrong.
AI isn't a magic box.
It's an incredibly powerful tool.
But like any tool, the quality of the result is ultimately determined by the person using it.
What do the established sources say?
Your experience lines up remarkably well with what major research and technology organizations are publishing today.
Gartner states that organizations with successful AI initiatives invest up to four times more in data foundations, governance and AI skills than organizations with less successful results. According to Gartner, successful AI isn't just about better models — it's mainly about reliable data and context.
IBM writes that “AI is only as good as the data it works with.” The company emphasizes that poor data quality is one of the main reasons AI projects fail, and that governance, representative data and human oversight are essential.
The concept of “Human-in-the-Loop” is getting more attention. It treats AI not as a replacement for human expertise, but as a tool that specifically needs human judgment and domain knowledge to produce reliable decisions. That maps almost one-to-one onto your own experience: you have to keep challenging AI, correcting it, and feeding it context.