Has anyone had to deal with applicants obviously using AI during interviews?
My company is in the process of hiring a Cisco network engineer with a minimum of 7 years experience. In the past, we have had interviewees who were obviously Googling answers during an interview. You could see them on cam stealthily typing or even reciting the question out loud so they could speech-to-text their answers. Unfortunately, it's getting harder to detect with AI integrations such as "Interview Co-pilot" which listens to the video call, searches for an answer on Claude, Gemini, and ChatGPT, and displays an answer.
I generally do the first round of interviews along with an HR rep to explain the specifics of the job and ensure they understand some of the unique responsibilities that the job entails. We had one particularly good candidate that answered some of my softball tech questions thoroughly and accurately. I sent her on to my lead engineers for a more detailed interview with troublehsooting scenarios and asking her to walkthrough a design approach for a specific network.
Initially we were very happy with the answers but since I had a backseat role in this interview, I noticed that the applicant was definitely reading answers from the screen. Even though the call quality was excellent, she would sometimes ask for a repeat of the question from the beginning. We asked a specific question about how a Cisco AP goes about finding the controller and registering and I already had the ChatGPT answer pulled up and it was 99% verbatim.
I was trying to find a question that would generate a hallucination from AI, but in the short period of time left, I came up empty-handed. When asked if she preferred CLI or GUI when configuring equipment, she said she mostly uses CLI, but will sometimes use SecureCRT to configure them. That's like asking if you fix your own car or take it to the shop and saying you mostly fix it yourself, but sometimes use a wrench to fix it.
The last question involved my engineer sharing his terminal window while logged into a switch. He displayed an access port and a trunk port with very specific commands on each port. The applicant was asked to review the ports and explain what each command does. This was the one time that they could not use AI to obtain their answers. It would have been too suspicious to read out all 8-10 lines and wait for a prompt, so they simply said "one is an access port, the other is a trunk port, what else do you need to know about them?" I am sure these AI apps will eventually be trained to read screens in the future, if not already existing in some way.
Has anyone had to deal with anything like this? I could screenshare all of our questions but I feel that could make for an awkward interview. One suggestion was to ask about a non-existent product or technical term or one that has nothing to do with Cisco networking (or networking in general) to see if they try to take the AI output and formulate a networking answer.