Today is 18 December 2025. None of us are Neo or Morpheus or Trinity or Link. Still, the agents are coming. Not physically. Metaphorically, if you happen to be an interviewer of a tech company who does a lot of online interviews.
While I was interviewing candidates for the open positions in my company, I came across a candidate who was using an AI agent for answering the questions in real time. He had passed two rounds and I was conducting the third round.
A colleague brought my attention to ParakeetAI - https://www.parakeet-ai.com/
Then I found ShadeCoder https://www.shadecoder.com/ and Interview Coder https://www.interviewcoder.co/
Obviously there are many tools and helpers out there. And some people are going to use those.
Then there is Talview https://www.talview.com/en/ and Canditech https://www.canditech.io/ and Fabric https://www.fabrichq.ai/
These are not for free. I didn't want to go in that direction.
Giving some thought to the situation, I wrote general guidelines for conducting online interviews and circulated them with my colleagues. Later I gave some more thought to this, and here is what I have so far.
1. The invitation of the online meeting should _not_ be open to all. The online meeting should be restricted only to the candidate and the interviewer.
2. After joining the online meeting, if the candidate keeps the camera off, ask the candidate to switch on the camera and be in a well lit place so that the candidate will be clearly visible during the entire interview. If the candidate refuses to switch on the camera for whatever reason, do not continue with the interview.
3. During the interview, check _where_ the candidate is looking when a question is asked. Is the candidate taking a pause, looking somewhere for an answer, and then start answering.
4. Check the authenticity of the resume. Ask the candidate about the things written in the resume.
5. Ask questions that an AI agent won't answer correctly. For example, when the candidate explains about the project he/she has done, ask "what was the most interesting defect that you fixed", or "what was the most challenging issue that you solved", or "what was the most interesting feature that you worked on". For questions specific to the candidate's past work, an AI agent gives generic answers which are not specific to any specific project, and hence can be detected.
6. Let's say you asked the candidate to write a piece of code or algorithm, or come up with a solution to some problem, and the candidate has written the code or the algorithm, or verbally explained a solution. Then ask the candidate to explain the code or the solution step by step. Or ask the candidate why he/she made certain choices. What would happen if some other choices are made.
7. Don't let the discussion go in a flow. Mix the questions.
8. Have open ended discussions, rather than a series of questions that would have known answers. Check the candidate's responses in the open ended discussions.
9. Show a piece of code that has problems or bugs and ask the candidate to spot those.
10. Focus the discussion around problem-solving, communication, and adaptability of the candidate.
11. A face-to-face interview in a physical presence is preferred, and should be done whenever possible.
If I were to live in The Matrix, I'd definitely be on the side of The Architect. More specifically, I'd be on the side of A. Smith. But in the world that we are (apparently) living in, we have to find ways to detect the agents, and be as humanly as possible.
And then there could be a different thought. During our regular day-to-day work, we are encouraged to use CoPilot. And using CoPilot is helpful in many situations. Then why not allow the candidates to use AI during the interviews.
Here is my view on this.
Although we allow supplemental use of CoPilot in our regular work, it is merely a helping hand that makes our job easy in some situations. It can not be used as the major skill. The skills and expertise an engineer collects during the course of his/her jobs and education is of value. That is what we are paying for. We are not looking for humans who use AI tools to do the jobs without being an expert in the area of work.
If a candidate announces upfront that he/she wants to use AI during the interview, or asks if AI is allowed during the interview, that is an honest way. Not announcing the use of AI and using it is cheating.
Should the company explicitly mention beforehand, "although using CoPilot on-the-job is encouraged, use of AI tools is 'forbidden' during the interview process"?
I am of the opinion that no such statement is necessary. We are paying for the skills and expertise of the candidate. Not the ability to use an AI tool. So in my view, it is reasonable to reject this candidate for his/her use of AI during the interview.
