How is AI changing real-time troubleshooting?

Before AI coding agents, troubleshooting a customer issue often meant cutting calls short, reproducing steps later, and long back-and-forth cycles. Now, AI tools like Devin (the AI software engineer from Cognition Labs) can help engineers resolve problems live during support sessions, improving efficiency and customer satisfaction.

What can Devin AI do in a customer support scenario?

A big part of my job as a developer is helping customers debug tricky issues.

The usual flow looks like this:

1. The customer hits an error and shares it with us. 
2. We go offline, dig into the code, try to reproduce the issue. 
3. One or two days later, we send back a curl command to help troubleshoot or find a workaround.

But recently, I had Devin AI open during a live customer call to help with the troubleshooting.

The customer showed me the error, and instead of taking it offline, I asked Devin for help—live.

It read through the code, figured out how the introspection API worked, and generated a working curl command—in a different language, no less. All in seconds.

The customer ran the command while we were still on the call, and it worked. We could immediately determine the root cause on the customer’s infrastructure side and develop a clear plan to fix it, right then and there.

How does Devin AI help troubleshoot problems faster?

Devin combines two powerful capabilities:

1. Indexed codebase: It can quickly locate and understand relevant logic.
2. Cross-language translation: It can convert code (e.g., TypeScript) into other formats (e.g., curl commands) in real time.

Together, these features enabled us to troubleshoot the issue live with the customer, without delays or multiple follow-ups.

This experience highlights how AI can:

• Reduce friction in customer troubleshooting.
• Save engineers time by automating command construction. 
• Improve customer confidence through real-time problem solving.

In this example, we provided faster, smarter support—turning frustrating sessions into opportunities to impress customers. 

Video walkthrough

Here’s a video walkthrough of how I used Devin AI for troubleshooting:

Full transcript: Using Devin to troubleshoot customer requests in real time

“One of the things that we sometimes have to work on as software engineers is customer support. And I want to show you one cool thing that I did with Devin that helped me a lot during a customer call.

We were trying to troubleshoot why a sync command was failing for a customer. And you know, in the pre-Devin world, we would have had to probably cut that meeting short and have a lot of back and forth while we figured out specific steps for them to reproduce the issue and try to isolate it.

So while we were in the meeting, I just went to Devin, and I just told it to give me a curl command to introspect the graph and to put placeholders for API URL key and graphs so that we could substitute it with the customers. The only hint that I gave it was that that functionality lives in the Faros JS client. So it's actually in this function here.

So imagine trying to construct a curl command based on this logic—you have to see what this does and then see what's in data and build the client schema and whatnot. That would take a good chunk of time. Certainly cannot be done live. But in a matter of seconds, Devin came up with an equivalent curl command that is equivalent to the logic that I was trying to test. We were able to test this live with the client without spending too much time and any back and forth.

I think this is one of the very powerful features that Devin has because it has all of your codebase indexed, but it's pretty amazing that it's also able to translate between languages. It's translating from TypeScript here to an actual crawl command.”

How to Use Devin AI for troubleshooting: Transform support from reactive to real-time

This approach didn't just save us hours—it fundamentally changed how we handle developer support challenges. What could have been a frustrating multi-day back-and-forth between teams became a productive, collaborative troubleshooting session that resolved issues in real-time.

I publish my thoughts on AI and experience with AI coding tools frequently. Follow me on LinkedIn for more tips on using AI coding agents.