Interviews in the AI age

In the era of AI, interviewing also has evolved significantly. These days, teams are worried both about what to evaluate during interviews.

When it comes to AI itself, teams are experimenting with several approaches. Some mandate NO AI. Others, like Realfast, who’ve always been generous with sharing their learnings), now give take home projects and what you submit is not just the output but the JSONL files containing your conversations you had with the AI agent. Another AI agent then evaluates the conversation & grades you on parameters they’ve tried to codify.

Recently, I had a chance to evaluate one such agentic interview process by a startup in the recruitment space. My takeaway: The results of the agentic interview process, at least with the small sample set I saw, did NOT correlate with the actual aptitude I’d observed in the volunteers. Metrics the AI agent used were vague & not provable.

My takeaway: It is quite straightforward to assert if someone is using AI badly. But ‘good’ usage is not as easy. I do NOT think AI is dependable in highlighting good candidates. At best, it should ONLY be used as a gate (at a very early stage) to remove the very incompetent.

If you are an interviewer interviewing freshers, here is my recommendation instead: fall back to the basics.

• Stop asking textbook & DSA questions. The ability to memorise is NOT a valuable skill anymore.
• Review their projects, introduce challenges in the work they’ve done & evaluate how they think. Given a problem, are they able to break it down, think through the solution & communicate it clearly. These are the skills that will matter in the days to come.
• Ask them about their lives, evaluate for drive, discipline, grit, analytical abilities, attitude & communication skills.
• Don’t treat all candidates as equal: Look at where they started from. A privileged candidate with excellent English & another who may not speak English as well because he’s learnt the language after schooling should NOT be evaluated using the same standards. The former may be immediately productive but the latter may probably be a more valuable member of your team in the long run.

If you are a fresher who is appearing for interviews, how to prepare is a more difficult decision. Every interviewer is different and unfortunately many will still evaluate you on aspects I think are NOT valuable.

Here are some things you can do to ensure the interview showcases your strengths.

• Do more than just the mandated projects your college asks you to.
• Find like-minded friends & come up with ideas together. Then execute them. This will give you real-world experience that will help you in your job search AND career. If you are lucky & your project has real world value, you & your friends may never even need to apply. Create your own incubators if your college does not have one.
• Find peers from other streams & work together to solve challenges across disciplines.
• Do the DSAs & the puzzles. Contribute to open source regularly. The point here is to show discipline & consistency. Personally, a high DSA score is not as impressive to me as a long streak.
• Avoid showcasing thin AI projects where you only wrap a call to an agent. These do not showcase your own skills in any way.
• Do some projects the hard way - without AI coding for you. Search for answers - this is a skill you will need to develop that AI is taking away from you by providing answers too easily.
• And last, something very tangential to AI but important: don’t take the interview too seriously. Often, this attitude sabotages rather than help. Instead, tell yourself that you will do your best but don’t care about the outcome. (Hmm, I wonder if the verse Karmanye vadhikaraste Ma Phaleshu Kadachana was trying to teach us this lesson too.)