AI phone screening for applicants: how it works and where it stops
A transit company that hires drivers in several cities gets more applications than its office can phone screen. The screener we built calls every one of them, usually within the hour, and asks the same questions every time.
An AI phone screener calls each applicant shortly after they apply, asks a fixed list of questions, writes up the answers, and recommends hire or pass with its reasons. Everything is logged to the hiring pipeline for a person to read. It never makes the decision and never tells an applicant they got the job.
Why phone screening was the bottleneck
For most of the businesses we work with, the hard part of hiring is not finding applicants. It is calling them back. Applications arrive at all hours. The office is busy. By the time someone calls, the good applicants have taken another job and the rest have stopped answering. A screener that calls within the hour, every time, changes who ends up in the interview chair.
What the call sounds like
The applicant gets a call from the company's number. The screener introduces itself by name, says it is an automated assistant helping with the first step, and asks if now is a good time. If not, it offers to call back later and does. If yes, it runs through the questions.
The questions are the same for every applicant for a given role, written with the hiring manager before the screener goes live. For a driver role they are the practical ones:
- Do you hold a valid license, and for how long?
- Which of these shifts can you work?
- How far are you from the location?
- Have you driven passengers professionally before?
- Is there anything on your driving record we should know about?
- When could you start?
It listens, asks a follow up when an answer is unclear, and moves on. It does not improvise new questions, does not small talk its way into territory the company has not approved, and does not ask anything the law says it cannot. Every applicant gets the same call, which is better for fairness than the version where a tired office manager asks whatever comes to mind.
What happens after the call
Within a few minutes the applicant's record in the hiring pipeline has three new things attached: the recording and transcript, a structured summary (each question with the answer in a sentence), and a recommendation of hire, pass, or talk to them, with the reasoning spelled out. "Available for the weekend shifts the location needs, has driven passengers before, lives close. Recommend interview." Or: "Cannot work any of the open shifts. Recommend pass, but keep on file."
The hiring manager reads that instead of making the call. If they agree, one tap moves the applicant to the interview stage and sends a scheduling link. If they disagree, they move it the other way, and the screener does not adjust itself based on that one decision. We do not let it tune itself on individual outcomes, because a few unusual hires would skew it.
Where it stops
This is the part we are strictest about, and the part owners ask about first.
- It never decides. A recommendation is a recommendation. No applicant is rejected or advanced without a person moving the record.
- It never tells the applicant an outcome. It says a person from the team will follow up. It does not say "you're hired," "we'll pass," or anything that could be heard as either.
- It never asks a question the company did not approve. No questions about age, family, health, religion, or anything else that does not belong in a screen. The list is reviewed by the owner and, when they want, by their employment attorney.
- It never pretends to be a person. It says what it is in the first sentence.
- It never handles the hard moments alone. If an applicant is upset, confused, or asks something it cannot answer, it apologizes, says a person will call, and creates a task for one.
Why the same questions matter
Consistency is the quiet benefit. When every applicant for a role is asked the same six things in the same order, the summaries line up and a hiring manager can compare a stack of applicants in a few minutes. Patterns show up: most applicants cannot work the early shift, or most live too far from one location. Those patterns feed back into the job posting and the schedule. A human screener could do this too, but in practice nobody writes it down the same way twice.
What the owner sees
A pipeline that is always current. Every applicant has been reached or has a documented attempt. The ones worth a human's time are sorted to the top with the reason. The ones who did not answer are retried on a schedule and then marked. Nobody is waiting on a call that was never made.
The screener is one of the agents we build around a client's hiring, alongside the applicant pipeline it writes into and the morning brief that tells the owner who is waiting. It makes the calls. The people make the decisions.
Questions people ask
Does the AI screener decide who gets hired?
No. It asks a fixed set of questions, writes up the answers, and recommends hire, pass, or talk to them with its reasons. A hiring manager reads the summary and moves the applicant. No one is advanced or rejected without a person doing it.
Do applicants know they are talking to an AI?
Yes. The screener says in its first sentence that it is an automated assistant helping with the first step, and it asks whether now is a good time before starting. If the applicant would rather wait, it calls back later.
What questions does the screener ask?
Only the ones the hiring manager approved before it went live, the same for every applicant in a role. For a driver that means license, shift availability, distance, experience, driving record, and start date. It never asks about age, family, health, or religion.
Want this done for your business?
Two minute intake. A real person reads every one and replies within a business day.
Keep reading
What an AI assistant on a website actually does all day
What a website AI assistant does all day for a Florida service business: page matched openers, tap qualifying, price callbacks, CRM leads and a morning recap.
AI at workHow we pick an AI model for a client project
How a Delray Beach studio picks an AI model for a client project: the job, speed versus depth, cost, privacy, and why one product often uses several models.
AI at workOpenAI, Anthropic, Google and the rest: what each is for when you are building
A builder's map of OpenAI, Anthropic, Google, Meta, xAI, DeepSeek and Mistral: what each is known for, hosted versus open models, and how we choose per job.