For most of my career, saying “I don’t know” has felt dangerous. We tend to reward the person in the room who has the answer “the expert”, the person who can immediately explain why something broke, how a system works, or what the team should do next. Nobody wants to be the person who says, “I don’t know.”

But I think AI is beginning to change that. In fact, I believe one of the groups that will thrive most in the age of AI will be what I call the “I don’t know” learners. These aren’t people who are comfortable remaining ignorant. They are almost the exact opposite. They are people who are comfortable admitting they don’t know something because their next instinct is to figure it out.

There are two very different versions of “I don’t know.” The first is, “I don’t know, so someone else will have to handle it.” The second is, “I don’t know. Let me figure it out.” I think that second person is going to be incredibly valuable in an AI-driven workplace.

Before AI, there was considerably more friction between not knowing something and becoming competent enough to do something about it. Maybe you needed to find the right documentation, track down a coworker with the right experience, search forums, watch videos, or piece together five different articles. Those resources were useful, but there was a significant cost to crossing the gap between “I don’t know” and “I understand this well enough to move forward.” AI is dramatically shrinking that gap.

I’ve always been someone who asks questions, and sometimes they’re pretty basic. I’ve asked, “What actually happens when I run “kubectl get ns”?” A simple question like that can lead to understanding authentication, networking, the Kubernetes API server, permissions, VPN connectivity, DNS, and why a command might fail before it ever reaches the cluster.

I’ve also asked, “If I have three GUI pods, how do I know which pod a particular customer is actually hitting?” That opens up a conversation about load balancing, ingress, sessions, logging, and observability. From there, I might wonder why nothing useful appears in the pod logs and whether the failure could be somewhere else. Now I’m learning about tracing requests across systems, network paths, and using tools like Dynatrace to figure out where a request actually failed.

I’ve had the same experience with automation. A question might begin with, “Can I automate updating these Kubernetes certificates with PowerShell?” That quickly becomes a larger problem: securely retrieving certificates, handling passwords, converting formats, generating Kubernetes secrets, validating the update, logging failures, and eventually figuring out how to safely perform the same process for 150 certificates instead of one.

That’s what I find powerful about AI. A single question can become an entry point into an entire system. Instead of searching for isolated answers, I can keep following the problem until I understand how the pieces connect.

There is, however, an important catch: if you don’t ask the question, you may never know there was something to learn.

AI can explain concepts, challenge assumptions, show relationships between technologies, and help troubleshoot problems. But it cannot magically give you curiosity. You still have to notice something and wonder why it works that way.

Imagine two people troubleshooting the same outage. One gets the application working again and stops. The other gets it working and asks, “Why did that fix it?” That question might lead to what actually failed, how the failure could have been detected earlier, whether it could be monitored, and whether recovery could eventually be automated. They both fixed the same problem, but one turned the problem into an opportunity to understand the system.

Sometimes, though, there’s an even harder problem: you don’t know what you don’t know. You may not understand a subject well enough to even know which questions to ask. This is another place where AI can help. Instead of only asking for answers, I can ask, “What questions should I be asking that I haven’t asked yet?” or “What would someone with ten years of experience notice here that I might overlook?”

I think that’s one of the most powerful uses of AI. It doesn’t just help answer questions. It can help us discover better questions. But we still have to initiate that process. We have to be willing to admit that there may be something we’re missing.

AI also removes some of the social pressure that has traditionally surrounded learning. Sometimes we don’t ask questions because we think we’re supposed to already know the answer. You’ve been in IT for ten years, so surely you should know this. You’re the senior person on the call, so surely you should understand that acronym. Everyone else seems to understand, so maybe you just stay quiet.

With AI, you can ask the embarrassingly basic question. You can ask for the same concept to be explained five different ways until something finally clicks. There’s no meeting to slow down and no fear that somebody is silently wondering how you got your job. You simply get to learn. That creates an extraordinary opportunity for people humble enough to admit the limits of their knowledge.

Of course, there’s a risk on the other side. AI can make it incredibly easy to get an answer without actually learning anything. If you ask AI a question, copy the response, and move on without understanding it, you’ve outsourced the work rather than expanded your knowledge.

That’s why I think the most effective relationship with AI isn’t simply, “Do this for me.” It’s also, “Help me understand this.” Challenge my assumption. Explain why this failed. Show me what I’m missing. Give me different approaches and explain the tradeoffs.

I also don’t believe AI makes knowledge irrelevant. Actually, I think the opposite may be true. The more you know, the better you become at recognizing when AI is wrong. Experience gives you context. Knowledge gives you judgment. Expertise helps you recognize when an answer sounds convincing but doesn’t actually make sense. AI can generate possibilities incredibly quickly, but somebody still has to evaluate them.

So the future isn’t knowledge versus AI. It’s knowledge + curiosity + AI. The person with twenty years of experience who refuses to learn anything new may struggle. The inexperienced person who blindly believes everything AI tells them will struggle too. But the person who combines experience with the willingness to say, “I don’t know—let’s figure it out,” becomes incredibly powerful.

Technology already moved quickly before generative AI, and now it feels like the speed has increased again. Tools change, platforms change, programming languages evolve, and best practices change. Entire categories of software can appear in just a few years. That means there’s an uncomfortable reality for all of us: at some point, every expert becomes a beginner again.

The question isn’t whether you’ll encounter something you don’t understand. You will. The question is what you do next. Do you protect the image of being the person who knows, or do you become the person who learns?

I’ve realized that I don’t need to know everything. I need to be willing to learn almost anything. That’s a very different mindset.

AI gives us access to something previous generations never had: a tool that can sit beside us while we’re learning, answer follow-up questions, explain unfamiliar concepts, critique our ideas, and meet us wherever our current level of understanding happens to be. That doesn’t eliminate the need for expertise. It gives curious people a faster path toward it.

So when I think about who will thrive in an AI-driven world, I don’t necessarily picture the person who walks into every room with the most answers. I picture the person who walks out of the room with better questions.

The person who isn’t afraid to say, “I don’t know.”

And then follows it with:

“But I can learn.”

Those are the people I’m betting on.