I completed an entirely AI-native screening process for a job yesterday. (Buckle up—it’s a brave new world out there.)

This process was so different (mostly in a bad way) from any hiring I've experienced in the past decade, I had to share—especially if you're considering AI in your own hiring.

AI-powered job screenings do have potential, but the extreme approach I encountered yesterday—a rigid mix of competency-based assessments and clinical extraction of details—completely sucked the humanity out of the process. To me, that felt hugely problematic.

In this blog post, I’ll walk you through the steps of this AI-powered screener, what I liked and where I saw things fall short, and include some considerations or thoughts for anyone out there who might be thinking about incorporating AI into your own hiring processes.

To back up, let’s start with a little context-setting as to what I did, and why this process felt so different to me from my normal job application flow.

What I did:
Here’s what I did: Caught in mid-afternoon YOLO energy, I quick-applied for an executive role at a stealth AI-native startup through the now-infamous (and infamously terrible) Crossover process. Normally, I’m not into this thing, but I was curious to see what an AI-first application process would feel like. So I decided, what the hell.

The Screening Components
The process involved three main components:

1. An AI chatbot-powered conversational interview

2. A cognitive assessment to test problem solving and computational thinking

3. A three-part prompt engineering assessment

Each part was live proctored, and if I navigated away from the browser, I got a slap on the wrist—this happened twice when my mom texted me in the middle of it. (Thanks, mom.)

Now let’s dig into each of the parts of the screening itself. 

Hot take: Fun idea, poorly executed. 

What happened: I had 30 minutes on the clock to answer questions from an AI chatbot to assess my overall competency level for the role in question.

Longer take: Having built AI-powered chatbots for interviews myself, I was already familiar with the process and therefore knew how to play the game. But I didn’t love the rigidity of their screening questions. For instance, I was asked for specific sales experience in a niche sector. While I have sales experience in one area and industry experience in another, this kind of nuance is much easier to explain in person than over a chatbot.

As someone familiar with LLMs, I suspect a less AI-savvy candidate might have struggled with this part of the screening. Even though I understood the process, the one-sided nature of chatbot conversation felt incredibly cold and clinical, giving no insight into the company's needs or an opportunity to ask questions in return or riff on ideas. Toward the end, I did try to “reverse interview” the chatbot, asking it how the AI was trained and what success looked like in an ideal candidate. It refused to answer any of my questions. Ah well. Worth a shot.

How to make this better in your recruitment process:

• Provide more upfront context and expectation-setting (like a recruiter would)

• Consider a dual-sided chat interface that encouraged questions from both sides (it’s hard to answer in a vacuum)

• Use a flexible, nuanced interview framework to avoid filtering out creative thinkers

• Include the option to voice dictate instead of typing out each response

• Add some personality to your chatbot, ideally reflecting your company’s brand and tone of voice

Hot take: Hard pass. Nothing makes you feel less human than a series of logic tests that most people fail the first time.

What happened: I had 15 minutes to complete 50 questions that asked me math questions, verbal questions, and problem-solving questions.

Longer take: Seeing as this whole process was more of a YOLO energy than as a serious application, I was not prepared for on the spot mental math, pattern recognition, and word games. I didn’t prepare at all for the assessment and did not have the right setup or pacing in mind when I started the first time. I failed with a 62% pass rate the first time through.

On my second attempt, at minute 10, I heard a blood-curdling screen from my kids’ bedroom and then my four-year-old ran upstairs to tell me that Sydney needed my help and her nose was bleeding. So much for pauses. So the last four minutes were spent with me panic attack filling in multiple choice answers while also triaging a bloody nose. Miraculously, everyone was fine. And I passed with 85% success that time. I guess I operate better under pressure.

How to make this better in your recruitment process:

• Just…don’t. No executive wants their first exposure to your organization to feel like an SAT test.

• But…if a test is mandatory, at least build in empathy for real-life interruptions. Speaking from experience, a “pause” button for those “bloody nose emergencies” would have been helpful.

Hot take: Super fun, but didn’t take it far enough.

What happened: I was given 40 minutes to solve three separate AI-powered challenges with what appeared to be a white-labeled version of Claude.

Longer take: This was the most enjoyable part of the experience for me. After six months of living and breathing through self-directed chatbot prompt challenges, it felt like an oddly fitting benchmark for my AI skills. It also clarified how natural it now feels to be assessed on my ability to work with AI—much like when my data skills were evaluated for a startup sales role. To me, this shift suggests that the future of hiring will likely center on demonstrating competency with the latest tools, especially AI.

Here’s how it worked: On one side of my screen, I’d see information like a meeting summary or a sample corporate workflow, and on the other, I was tasked with prompting the AI to solve the given problem. Each example was essentially a puzzle with a specific solution in mind. While that is obviously one way to use AI, it really didn’t lean into the generative possibilities of AI. Personally, I would have been more excited to receive a more open-ended prompt to assess my creative potential and ability to use AI to remix or transform content, ideas, or processes from one straightforward answer to something a bit more inventive. Still. Kinda cool.

How to make this better in your recruitment process:

• Instead of arbitrary AI tasks, encourage candidates to use AI to solve a relevant “take-home” assignment. (Just be sure to compensate for any take-home work.)

• Lean into seeing how candidates handle the iterative process of using AI, not just how well they can architect or design an initial prompt.

• Ask candidates to share the chat log as proof of their work.

• Go all in on creativity, or vary up your AI challenges with one accuracy-based challenge and one creativity-based challenge.

Are AI-powered hiring processes the future, or an overreach? Why not both?

For over a decade, strategic networking has been my primary path to new roles. I know this has its own bias—I don’t face the usual barriers to reaching decision-makers because of my network. The platform for this AI-powered process promoted its ability to “normalize” hiring by leveling the playing field and using a uniform rubric. I don’t doubt that it reduces certain biases; the whole experience felt less like my usual networking and a lot more like participating in a blind, sight-reading music audition behind a curtain, with no one seeing or knowing who I am.

On one hand, it’s true that AI screenings can bring objectivity to hard skills and aptitudes. But on the other, they risk introducing new biases—one that converts humans from complex and emotionally charged people with motivations, dreams, and fears, into task-achieving robots, defined only by our skill sets. The process didn’t allow me to convey the “squishier” aspects of my personality or the intangibles I tend to bring to group dynamics. And for me, those nuances—the places where I like to color outside the lines—are often the most interesting things to discuss in a job interview.

Truth be told, I’m still not convinced this whole process wasn’t just a clever way of tricking me into oversharing personal information online for a potential future scam. Wouldn’t be the first time I’ve fallen for a long-game phishing scheme. (If so, well played—and please ignore any bulk emails from me for the next few weeks.)

In the end, the eerie absence of any human touch left me wondering: Was this entire process run by an AI, with no human oversight? Did I just apply for a job to work for an AI boss?

Oddly enough, if that were the case, I'd respect this process a lot more.