Preparing to Switch Jobs as a Developer: Thoughts on the AI Interviewer Era
Recently, while preparing to change jobs, I noticed that companies’ approaches to using AI fall into a few categories. This is based on my experience, so it may be biased.
The first type is resume screening.
They match the JD keywords with your resume using an LLM for initial filtering. In this case, if you only write about good experience without keywords, it can just get tossed. I fed the JD of the company I was applying to into an AI, checked which keywords overlapped, and polished my resume using that phrasing. After doing this, my resume pass rate jumped significantly.
The second type is an AI competency interview.
They actually ask, “How do you use AI tools?” To do well here, you need concrete examples. When I described workflows like “before code review, I ask it to summarize the static analysis results and attach them,” they seemed to like that.
The third is just the same as before.
Half of places still have humans review everything and run technical interviews the old way. These days, some companies even don’t allow AI use. I don’t know if it’s a security policy or just because things haven’t changed.
In short, there are two tracks to prepare for. One is still the fundamentals (CS, algorithms), and the other is a concrete story about “how I use AI in my work.” The latter didn’t exist before, but now it’s almost mandatory.
One-line summary: Being good at using AI has become a credential. It sounds strange, but it’s true.
5 answers
Agreed. I especially resonate with #1. I also spent two months getting instantly rejected at the resume stage because I didn’t match a single JD keyword, but after I looked at the JD and just changed my wording, I started landing interviews right away. It wasn’t a skill issue—it was an optimization issue lol.
But in the third category, lumping “companies that won’t let you use AI” in with places that simply haven’t changed doesn’t seem right. There are plenty of projects where pasting source code into an external LLM is itself a contract violation. If it’s a security policy, they should just say it’s a security policy.
There are way more of the second type of question now. But if all you say is "I'm good with AI," everyone sounds the same, and it actually backfires. I also talked about my failures, and that got a much better response. For example, the time I merged generated code as-is and then had to roll it back.
Well, if their pass rate went up because of keyword matching, doesn't that mean the company evaluates candidates that shallowly? I think it's right for places that hire that way to get filtered out. Anyway, even if you get in, there's a high chance the work will be weird.
Got a source? I’m curious about that “half is still the old way” figure. From what I’ve seen around me, only the paperwork is AI, and almost everything else is still done by people.