The First Ten Seconds of an AI Screening Call

Since January 1, employers in Illinois have had to tell candidates when they use AI to make hiring decisions, under an amendment to the state's Human Rights Act [1]. Put that next to New York City's rule that candidates get notice before an automated employment decision tool is used on them [2], and the EU AI Act's requirement that people be told when they're dealing with an AI system [3], and AI disclosure in hiring is no longer optional in a lot of the places our customers recruit.

Most of the discussion so far has been about where the notice goes: the job ad, the privacy policy, a checkbox on the application. That matters to the lawyers. I think the disclosure that counts happens somewhere else. It's the first ten seconds of the call, when a candidate picks up and hears a voice that isn't a person. How you handle that moment decides whether the next eight minutes are a screen or a candidate trying to work out what's going on.

Candidates notice anyway

Some teams would still rather blur it. The reasoning I hear is that people hang up on bots, so let the call get going before anyone mentions it, or give the voice a human name and leave it there. I get the instinct. I also think it misreads how people actually experience these calls.

Voice AI is good now, but people who've spent two years fielding robocalls and scam texts pick up on small tells fast: a pause before an answer, a reply that's a little too tidy. When a candidate realizes halfway through that they're talking to software and nobody said so, the question in their head changes from "do I want this job?" to "what else haven't they told me?" We wrote last week about how job scams have taught applicants to distrust a stranger's first message. An undisclosed bot is exactly what a scammer would use, and candidates will treat it that way.

There's a quieter cost as well. A candidate who doesn't know they're talking to an AI can't make sensible choices about the conversation. They don't know whether to slow down, whether they can ask for a question again, or whether the odd answer they just gave will be read by a person with context or scored by a model without it.

What a good disclosure actually says

"This call may use artificial intelligence" is technically a disclosure. It also tells the candidate almost nothing. A good opening answers the questions they're about to have, in plain speech, because they're listening, not reading.

Something like: "Hi Maria, this is the screening assistant for Northside Logistics. I'm an AI, calling about the forklift operator job you applied for on Sunday. I've got about eight questions on shifts and experience, so roughly ten minutes. A recruiter on the team reads every answer and makes the decision. If you'd rather talk to a person, just say so and someone will call you back."

Five sentences, and none of them is filler. Naming the company and the job means it doesn't sound like a scam. It says it's an AI without apologizing for it. It tells them how long this will take. It says who decides, which is what candidates worry about most and the thing most often left out. And it offers a way out. Very few people take it, but knowing it's there changes how the rest of the call feels.

The line about who decides has to be true. If a score below some threshold rejects people automatically, don't tell them a recruiter reads every answer. Say what really happens. California's rules on automated hiring decisions assume you may have to explain your process years later, and what you told the candidate at the start of the call is part of that record.

The hard part is saying it every time

This surprised me when we started listening to real calls. Writing a compliant notice is easy. Making sure every candidate actually got it is not. People are inconsistent about this kind of thing in ways nobody notices: one recruiter explains how the process works, another skips it on a busy afternoon, a third mentions the AI assessment only if the candidate asks. If the obligation is to notify every candidate, a notice that depends on who was on shift that day isn't a policy.

So the practical version is short. Write the opening down. Put the same wording, or a shorter version of it, in the job ad and the application confirmation, so the call isn't the first time the candidate hears about it. Decide what the AI does and doesn't decide, and make the script match. Then keep proof that each person heard it.

How Asendia AI handles this

Asendia is a voice-first AI recruiter. It phones every applicant, usually within hours of them applying and at any hour, 24/7, and runs a structured screen for the role. You write the opening, and we push customers to make it specific: the company, the job, the fact that it's an AI, what happens to the answers, and how to reach a person.

Because the opening is part of the script, every candidate hears the same one, whether they're called at 2pm on a Tuesday or 11pm on a Sunday. Every call is transcribed, so if a regulator or a candidate asks whether someone was told, you can show them the exact words. The answers and a short summary go into the ATS you already use, alongside the rest of the file. Most agencies using Asendia came to us for volume, to screen hundreds of applicants a week without hiring more recruiters. Their compliance people usually ask about the disclosure first.

Final Word

Hiding the AI doesn't get you better candidates. It gets you the ones who didn't notice, plus a few who did and stayed on the line wondering what else you'd left out. The law is slowly catching up with something that's obvious from the candidate's side: people want to know who, or what, they're talking to, and who makes the call.

This week, listen to the first thirty seconds of five screening calls your team ran, human or AI. Write down what each candidate was told about how the decision gets made. If the five answers don't match, start there.

Ready to transform your hiring strategy? Schedule a Demo with our founders today!: https://asendia.ai/talk-to-founders