AI Phone Screening: What Should Recruiters Automate?

A job advert goes live on Monday morning. By Wednesday there are three hundred applications. Your recruiters read the shortlist, then spend the rest of the week on the phone asking the same six questions: are you available, where are you based, what pay are you looking for, do you have the right to work here, can you start on the date we need, and do you hold the licence. Most calls end in a no for a simple reason. The good candidates, meanwhile, wait two days for a call and accept somebody else's offer.

AI phone screening aims at exactly this work. It also sits in the part of the hiring process where the law is strictest and where a mistake can hurt real people. This guide is for recruiters, staffing agency owners and talent acquisition leads who want to know what to automate, what to keep human, and what the rules say.

Here is what matters most:

• Use AI for the logistics of the first call: availability, location, pay range, right to work, start date and licences. These are factual questions with clear answers.
• Do not use it to judge people. Scoring a candidate's voice, accent, fluency or personality is unreliable and risky, and it invites discrimination claims.
• A person should make every decision to reject or advance. The AI collects answers and books the next step.
• Hiring automation faces specific rules: New York City requires bias audits and notice for automated employment decision tools, the EU classes recruitment AI as high risk, and in the UK solely automated decisions with significant effects are restricted.
• Candidates must always have a way to reach a person and a fair alternative if the phone does not suit them.

What AI phone screening is, and the two meanings of the term

The phrase is used for two different things, and the search results mix them. In one, it is a feature on a consumer phone that screens unknown callers. In the other, which this guide is about, it is software that rings job applicants, asks a set of screening questions, records the answers and passes the results to a recruiter.

In hiring, you will also see it sold as automated candidate screening, an AI recruiter, an AI phone interview or recruitment automation. The details vary. The common core is a voice agent connected to your applicant tracking system, the software where applications and candidate records live. When someone applies, the agent rings, or texts to arrange a time, runs through the questions you wrote, writes the answers into the candidate's record and, if you wish, offers interview slots from the recruiter's calendar.

It differs from candidate screening software that scores CVs or ranks applicants. That software reads documents. A phone screening agent has a conversation. The two are often combined, and they carry different risks.

What is worth automating

The best candidates for automation are the calls where the answers are facts and the questions are the same every time.

Availability and logistics. When can you start? Which shifts can you work? Do you have transport? Can you travel? These questions decide whether the role is even possible, and they take a recruiter minutes each time.

Location and right to work. Where are you based, and are you legally able to work in this country? The agent can ask and record the answer. Recruiters should verify documents themselves.

Pay and notice period. What salary or hourly rate are you looking for, and what is your notice period? Useful for filtering mismatches early, and kinder to candidates than a wasted interview.

Licences and certifications. Do you hold the licence or certificate the role requires? In high-volume hiring for drivers, nurses, electricians or security staff, this is often the first filter.

Scheduling. Offering the candidate a choice of times for the interview, booking it in the recruiter's calendar and sending a confirmation. This single step often saves a surprising amount of back and forth.

Reminders and no-show reduction. A call or text the day before an interview, with an easy way to reschedule.

In high-volume hiring such as warehouse, hospitality, care, retail and seasonal work, this is a large share of recruiter time. It is also the part candidates dislike least, because it is quick and arrives soon after they apply.

What to keep with a person

Equally important is what the AI should not do.

Judging fit. Whether a candidate will do well in the job, get on with a team or suit a client is a human judgement. An AI cannot read it reliably, and pretending otherwise is a risk.

Scoring voice, accent or fluency. Some tools claim to grade how a candidate sounds. I would avoid them. Accent and speech patterns correlate with national origin, disability and age, and decisions based on them are hard to defend.

Rejecting candidates. The AI collects answers. A recruiter should decide who goes forward, and should be able to see why. A candidate rejected by software without anyone checking is the situation most hiring rules target.

Sensitive topics. Health, disability, family plans, age, religion and similar subjects should not be raised by the AI at all, and it should not record them if a candidate volunteers them. Build this into the instructions, and test it.

Anything the candidate finds distressing. If a candidate becomes upset or asks to speak to a person, the agent should hand over straight away.

A short example shows the shape. Imagine Priya, who runs a staffing agency supplying warehouse operatives. Applications arrive overnight. By seven the next morning an AI agent has rung or texted each applicant who met the basic criteria, asked about availability, location, right to work and forklift licence, and booked those who passed into interview slots on her recruiters' calendars. The recruiters spend their day talking to people who are already qualified and available, not working through a list of calls that mostly end in a no. Nobody was rejected by the software: candidates who did not meet a hard requirement were told clearly, and anyone could ask for a person.

You can hear a booking conversation in the live demo below. It is one of our own demo agents, set up as a salon booking agent, and it handles a similar scheduling task. Try interrupting it, changing the time and asking something unexpected.

The rules you need to know before you start

Hiring is a regulated area, and the rules are moving. I am not a lawyer, and the details change, so use this as a list of questions for a lawyer who knows employment and technology law where you recruit.

New York City. Local Law 144 covers automated employment decision tools used for hiring or promotion in the city. Employers and agencies must have an independent bias audit within the previous year, publish a summary of the results, and notify candidates in advance that such a tool will be used. Whether a particular phone screening tool counts depends on how much it influences decisions, which is another reason for it to collect answers and leave the decision to a person. Check the current rules and notice timing.

The European Union. The EU AI Act treats AI systems used for recruitment and candidate evaluation as high risk, with obligations on providers and on employers who use them, including human oversight, record keeping and informing candidates. The dates for those obligations have been under revision, so check the current timetable. The requirement to tell people they are interacting with AI is separate.

The United Kingdom. UK GDPR restricts decisions based solely on automated processing that have legal or similarly significant effects on a person, and a hiring rejection can be one. Keep a person in the decision, tell candidates what data you collect and why, and keep records.

Phone consent in the US. An AI voice counts as an artificial voice under the Telephone Consumer Protection Act, so calls to candidates' mobile phones need consent. Capture it on the application form, in plain words, saying that you may call and text using automated technology. Our guide to AI outbound calling explains the rules in detail.

Discrimination law generally. Whatever the technology, hiring decisions must not disadvantage people on protected grounds. In the US a rule of thumb known as the four-fifths rule compares selection rates between groups to flag possible adverse impact. Track pass-through rates by group on any automated step, and look into any gap.

Candidate experience: the part that decides whether it works

Candidates talk about how they were treated, and a bad automated call damages your employer brand or your agency's reputation.

Say at the start that it is an automated assistant, who it is calling for and how long it will take. Keep it short. Ring at a time the candidate has agreed, which is another reason to text first. Let them reschedule easily. Offer a way to speak to a person at any point. Provide an alternative for candidates who cannot use a voice call, such as people who are deaf or hard of hearing, have a speech impairment or are not comfortable in the language. Tell candidates what will happen next and when. Nothing is worse than a screening call that is followed by silence.

Also let candidates know how their data will be used and for how long it will be kept, and delete it when it is no longer needed.

Connecting it to your tracking system, and measuring it

Connect it to your applicant tracking system

An AI phone screen that leaves its results in a transcript has saved no time. The answers need to land in the right fields in your applicant tracking system, such as Greenhouse or Bullhorn, so recruiters can filter and act on them. That means connecting the voice agent to the system, reading the application, writing the answers back, updating the candidate's stage and booking the interview in the right recruiter's calendar.

The conversation is the easy part. The work is in making sure the agent identifies the right candidate, writes to the right record, does not duplicate candidates and does not overwrite a recruiter's notes. This is the engineering we do in our AI voice agent development projects. For a wider view of where voice agents work and where they fail, see our guide to AI voice agents for business.

Measure speed, quality and fairness

Track speed, quality and fairness. For speed: the time from application to first contact, and the share of qualified candidates reached within a day. For quality: the share of screened candidates who pass the recruiter's review and the interview, and recruiter hours saved. For candidate experience: completion rate, requests to speak to a person, and how candidates rate the experience. For fairness: pass-through rates by group at each automated step, reviewed regularly.

Pilot on one role or one location. Listen to a sample of calls, check what the system wrote to the records, and compare the results with your manual process before widening.

FAQ

What is AI phone screening?

In recruiting, it is software that rings job applicants, asks a fixed set of screening questions such as availability, location, pay and licences, records the answers and passes them to a recruiter. It can also book interviews. It is meant to handle the repetitive first call, leaving recruiters to assess and decide.

Is AI phone screening legal?

It can be, with care. The law does not ban it, but hiring is regulated: some places require bias audits and notice, the EU treats recruitment AI as high risk, and the UK restricts solely automated decisions. In the US, calls to mobile phones need consent. Keep a person in every decision, and take advice for the places where you recruit.

How should candidates handle AI phone screening?

If you are a candidate, treat it like a recorded call. Answer the factual questions clearly, speak at a normal pace and ask for a person if you would prefer one. Employers should always offer that option. If a process seems to reject you without any human involvement, ask how decisions are made.

Can an AI recruiter replace a human recruiter?

For the repetitive parts, such as first calls and scheduling, software does a good job. For judging fit, persuading a hesitant candidate, advising a client and handling the delicate conversations, people remain essential. A realistic outcome is recruiters spending more time on the work that needs them.

What is the best candidate screening software?

There is no single best one. Judge any option on how it connects to your applicant tracking system, what it asks and records, how a person reviews its output, what testing for bias it supports, and what it does when a candidate asks for a human. Software that makes the decision for you is the wrong choice in most jurisdictions.

Does AI phone screening work for high-volume hiring?

That is where it helps most, because the questions are the same and the volume is large. Warehouse, hospitality, care, retail and seasonal hiring all fit. It matters even more to reach candidates quickly in those markets, since good candidates often accept the first offer they get.