The tools we use in early-stage screening are evolving fast. And not always for the better.
For years, hiring teams leaned on one-way video interview software to help them scale screening. Candidates recorded answers on their own time. Recruiters reviewed them later. Then came the AI wave; now we’re seeing tools that don’t just record responses but conduct interviews themselves, using conversational AI to ask questions, interpret answers, and even score performance in real time.
On the surface, it sounds like progress. But does agentic AI in recruiting actually deliver a better experience for the candidate—or better signal for the hiring team?
We dug into both sides of the experience and the design implications that come with each. Here’s what we found.
The AI agent interview are less human and create more friction
In theory, a voice-based AI interviewer should feel more personal than a one-way video. In practice, it can go sideways fast.
Imagine this: A candidate finds themselves face-to-voice with an AI agent. The job description itself had promised an AI-forward culture. What they didn’t expect was that AI would be the first interviewer, and not a very good one.
1. Unclear expectations created stress
The interview invite was vague. No timing. No structure. Just “expect to hear from your AI interviewer soon.” When it did arrive, it didn’t clarify much. Should the candidate prepare for behavioral questions? A resume deep-dive? A technical case?
That uncertainty created a burden for the candidate to over-prepare and mentally brace for the unknown. Instead of setting the tone, the system left them guessing.
2. Automated summarization ≠ real listening
Throughout the interview, the AI agent paused only to “summarize” answers; jargon-filled, cookie-cutter recaps that sounded vaguely affirming but added zero value. The tone was upbeat, almost human. But the content was shallow.
Instead of feeling heard, the candidate felt interrupted and “mansplained” by a machine. Complex responses, like aligning eight department heads around a design rollout, were flattened into three bullet points.
AI has no working memory. It can echo. But it cannot understand.
3. Human pacing was ignored
Deep, open-ended questions came rapid-fire. Pauses triggered transitions. The AI agent didn’t wait. It didn’t adapt. If the candidate stopped to think, it would jump in with the next question, regardless of whether the answer was done.
The candidate had to fight to finish thoughts. They even found themselves speeding up, just to avoid being cut off. The result? Shallower answers and a loss of nuance.
As they later reflected: “It felt like I was the one who needed to act like AI: quick, efficient, always ready. But I’m human. I need time to think.”
4. When things break, no one’s home
Twice during the interview, the system cut out. There was no live support. No real-time troubleshooting. Just silence.
They had to follow up, wait hours for a response, and hope someone was on the other end.
What one-way video interviews get right
Compared to the AI agent experience, one-way video interviews are refreshingly straightforward. Candidates know they’ll be presented with questions. They have control over when to start. They get space to think and structure their answers.
What’s missing in both formats, of course, is real-time back-and-forth. But one-way video doesn’t pretend otherwise. It doesn’t try to simulate empathy or fake feedback. And that makes it feel less manipulative.
Importantly, asynchronous interview tools also don’t interrupt candidates. They’re built around the idea of thoughtful, self-paced responses. And the best of them allow for retries or practice sessions that ease candidate anxiety.
For hiring teams, the benefits are clear:
- Consistency in question delivery
- Time-shifted reviewing with highlights or AI-assisted summaries
- Scoring or matching frameworks that can be customized or standardized
- Reduced bias when paired with structured rubrics
Done well, one-way video interviews offer a scalable way to capture signal without compromising candidate dignity.
So which tool should we choose?
It depends on the role, the team, and the experience you want to create.
Use asynchronous video if you want:
- A low-friction, low-risk screener for early-stage roles
- Control over question format, time limits, and retries
- Human review with structured matching
- A more natural-feeling experience for candidates
Use AI-led agents if you need:
- Conversational data for language-heavy roles (e.g. sales, support)
- Real-time routing or escalation (e.g. call center pre-screens)
- High automation at massive scale; with a plan to monitor performance and error rates
- A clear strategy for human oversight and transparency
What we don’t recommend: using AI agents to simulate a thoughtful human conversation when they can’t actually deliver one.
How to design a better first-round screen
Whether you use AI-video interviews, one-way video, or even human phone screens, the principles of good design remain constant:
- Communicate clearly. Don’t leave candidates guessing. Set expectations upfront.
- Respect their time. Offer scheduling flexibility and transparent time commitments.
- Give space to think. Build in pacing that mirrors real human conversations.
- Don’t fake empathy. Summarizing someone’s answer is not the same as listening.
- Plan for the ‘unhappy path.’ Technical issues happen, so design your response accordingly.
The most effective tools in hiring aren’t always the most advanced. They’re the ones that help us make better decisions without breaking the trust of the people we hope to hire.
Start with the human, not the tech
It’s tempting to reach for whatever’s newest. But in hiring—especially for high-touch or leadership roles—speed isn’t the only metric that matters. Clarity, respect, and signal depth are equally important.
AI agents may someday augment how we interview. But they are not a shortcut to better hiring outcomes, at least not yet. Until they can listen with nuance, adapt to human rhythms, and convey genuine feedback, they will remain what they are: tools. Not teammates.
Let’s design accordingly.
