share

I was driving when a small orange light suddenly appeared on my dashboard. Check engine.

For a second, I waited for something else to happen. A strange noise. The car slowing down. Anything that would tell me what was actually wrong. Nothing did.

The car kept moving exactly as it had before. But now I couldn't stop looking at that light. Was it something minor? Was something about to fail? Could I keep driving?

And that was the frustrating part: the car knew enough to tell me there was a problem, but not enough to tell me what the problem actually was.

The light could tell me what was happening. It couldn't tell me why. I didn't think much about that drive afterward. But during my internship at Smartlinks, I started seeing the same problem in a completely different place: supply chains.

 

 The Warning Light in a Supply Chain

A delayed shipment isn't all that different from that check-engine light. A system can tell you that a shipment is late, but that's only the warning. The more important questions come next. Why is it late? Is it the carrier, an appointment, paperwork, the dock, or something else entirely?

And what happens next? A delay rarely stays contained to where it started. It can affect delivery promises, increase costs, interfere with other loads, and eventually reach the customer. The delay is what you can see. The harder part is understanding what caused it and what it might affect next.

That's something I saw much more clearly while working on TRUKR, Smartlinks' Transportation Management System. Smartlinks describes itself as an AI-enabled supply chain execution platform building toward a broader vision of an autonomous supply chain, where technology can recognize what's happening, understand its impact, and help people respond with less manual intervention.

The part I experienced was one piece of that larger vision, but the idea behind it was the same: Don't stop at the warning. Understand what's underneath it.

 

From Answering Questions to Finding Problems

I saw this idea take shape in two ways. The first was a chatbot clients can use to ask something like "where is this shipment" or "why is this load delayed" directly, in plain language, instead of digging through dashboards themselves to find the answer.

The second was something I got the opportunity to explore myself: an idea we've been calling an RCA, or root cause analysis, agent. Instead of sitting in a chat window waiting to be asked, it would run in the background, watching for exactly this kind of problem before anyone goes looking for it.

The chatbot waits for someone to ask a question. The RCA agent wouldn't have to.

Imagine a shipment that should be getting loaded but isn't. Instead of waiting for someone to notice, search for the shipment, and investigate what happened, the idea is for AI to recognize the issue earlier and start connecting the dots: why isn't it moving, what's likely causing it, what could it affect next, and who needs to know.

It brought me back to that check-engine light. Knowing there's a problem is useful. Understanding why it's happening is what makes the warning actionable. The goal isn't for AI to make every decision. It's to help people move from what happened to why it happened, faster. What I didn't expect was that Smartlinks would teach me to use AI in almost exactly the same way.

 

My Own Check-Engine Light

When I joined Smartlinks, I'd worked with GitHub and codebases before, but never anything at this scale. I was moving between multiple repositories, trying to understand code I hadn't written, and figuring out how different pieces connected. At first, the amount of information could be overwhelming.

My own warning light was pretty simple: I don't understand this. And Claude made it incredibly easy to make that warning disappear.

I could give it the problem, get an answer, make something work, and move on. But getting something to work and understanding why it works are two very different things. So I started using AI differently.

There were days when I'd spend almost the entire day going back and forth with Claude on a single concept. I'd ask question after question, explain what I thought was happening, let it challenge or correct my understanding, and then ask again.

Why is this function here? Where is this data coming from? How does this connect to another repository? Is my understanding of this right? What am I missing?

Sometimes one answer created three more questions. Sometimes I'd explain my understanding and Claude would show me exactly where I had gone wrong. Other times, the thing I thought I didn't understand wasn't actually the problem at all. There was another concept underneath it that I needed to understand first.

In a way, Claude became less like something I went to for answers and more like a consultant I could keep questioning. A consultant can simply hand you the recommendation, or they can walk you through how they reached it. One gives you an answer. The other leaves you with an ability.

I wanted the second. I wasn't trying to make my warning light disappear. I was trying to understand why it had turned on. 

 

Building the Foundation

That experience changed the way I thought about learning. If I accepted an answer from Claude that I didn't understand and kept moving, I might finish the task in front of me. But the next concept could depend on the one I had skipped. Eventually, I'd be building on gaps.

Understanding became my foundation.

So when I found a gap, I tried not to build around it. I worked through it first. It wasn't always faster. There were days when I spent much longer understanding something than I technically needed to finish the immediate task.

But over time, I could feel the difference. The repositories became less overwhelming. I started recognizing connections faster, and concepts that once took hours to understand became easier to reason through.

AI hadn't removed the learning. It had helped me build the foundation faster.

 

Becoming AI-Native

That's probably the biggest thing Smartlinks changed for me. Before this internship, I thought being good at using AI mostly meant knowing how to get better answers from it. By the end, I had started thinking about it differently.

Something like the RCA agent shouldn't force someone to manually search through dashboards just to understand why a shipment isn't moving. AI can remove that unnecessary search and help surface the right information earlier.

I learned the same lesson personally. I didn't need to spend hours searching through thousands of lines of code just for the sake of doing it manually. Claude could help me find what mattered faster. But once it got me there, I still needed to understand what I was looking at.

AI should spare you the search. It shouldn't spare you the understanding. Because understanding is the foundation everything else gets built on.

That's what becoming AI-native started to mean to me at Smartlinks. Not handing my work over to AI, and not refusing its help either. It meant knowing how to use AI to remove unnecessary work, challenge my thinking, ask better questions, and ultimately make my own understanding stronger.

I started the internship seeing AI as something that could help me get to answers faster. I left realizing that sometimes its bigger value is helping me get to the right questions faster. And looking back, that's what that little orange light was really asking me to do in the first place.

Don't stop at what. Ask why.

ESHA RAI INTERN

Author

We work faster than
you can even imagine


WhatsApp