"Can AI screen resumes?" Yes, it can read a pile faster than you can. It can summarize each resume, line it up against the job you wrote, and list the people who look worth a call. It should not decide who is rejected, and it should never be the only reader. This article is for the owner of a company with ten to two hundred people who hires a few times a year, gets more applications than anyone has time to read, and wants to know what is safe to hand over.
I was COO and then President of a public company that closed more than forty acquisitions, and I ran the integration on most of them. I am a lawyer by training. Now I write fixed-fee AI plans for CEOs of private-equity-backed and owner-led mid-market companies through 013 Labs. This is not legal advice. Hiring rules differ by country, state and city, so ask your own adviser before you rely on any tool to filter people.
What a tool does well
Reading the pile. A hundred resumes, each summarized in a few lines: where the person worked, for how long, what they did. The part nobody enjoys.
Matching against what you asked for. You write the must-haves and the nice-to-haves. The tool shows which resumes meet each one, and where it could not tell.
Pulling out the facts. Dates, job titles, licences, certifications and locations into one tidy list you can sort.
Drafting the messages. An acknowledgement to everyone who applied, a polite note to those you pass on, an invitation to a first call. You read them before they go.
Preparing the interview. Questions based on the job and on gaps in a resume, so the conversation starts from the person and not from a blank page.
Tidying the record. Who applied, when, what happened next, in one place. Useful if anyone ever asks why a person was not chosen.
What a person must keep
The criteria. What the job needs is your decision. Write it down before any tool sees a resume. A tool matching against vague criteria will produce vague results.
The rejections. A tool can be wrong in ways nobody notices: a good person with an unusual career path, a gap that has a perfectly good reason, a resume written in a different style. A person looks at everyone the tool would turn down, at least until the tool has earned some trust.
The interview. You are judging whether someone will do the work and fit your team. That takes a conversation.
The offer and the decision. Who is hired, at what pay and on what terms belongs to a named person in your company.
The fairness check. Anything that treats people differently because of age, sex, race, religion, disability or similar traits is a legal and human problem, and the company is responsible even when a tool did it.
The risks that are specific to hiring
It can copy old patterns. A tool learns from examples. If the past hires were all of one kind, it can quietly favour that kind. You will not see this unless you look.
It can reward the wrong thing. A resume that is well written by an assistant can look better than a plain one from a better candidate. Neat is not the same as good.
It can be fooled. Some candidates now paste your job post into an AI tool and tailor every line. The resumes can look more alike than they are.
Some places have rules. A few cities and states require notice to candidates, or an independent check, when an automated tool helps decide who moves forward. New York City is one example. Find out what applies where you hire.
Candidate details are personal data. Resumes hold names, addresses, work histories and sometimes more. They go only to a provider your company has approved, on a company account, with terms you have read. Is ChatGPT safe for business use explains what to check.
Starting safely
Start with summaries, not scores. Ask the tool to summarize each resume against your criteria. Do not ask it to rank people or give a number, and do not let it auto-reject.
Write the criteria first. Must-haves and nice-to-haves, in plain words, tied to the work and not to a type of person.
Read a sample yourself. Take twenty resumes, read them yourself, then compare with what the tool said. Where you disagree, find out why.
Look at the people it would pass on. Someone in your company reads those, every time, before a rejection goes out.
Tell candidates. If a tool helps with the first read, say so in the job post or the application. Some places require it, and it is the honest thing to do.
Keep a record. What the criteria were, who looked, what was decided. Do I need an AI policy for my small business shows how to put the rule on one page. The hiring steps are worth writing down too, and can AI write standard operating procedures covers drafting them.
Picture it going wrong
A company has two hundred applications for one role and switches on a tool that scores each resume and drops the lowest half. Nobody writes down the criteria. A strong candidate who changed careers scores low and is never seen. Another candidate hears nothing for weeks. Months later, someone asks why the shortlist looks so alike, and nobody can explain how it was made. None of it needed a bad tool, only missing criteria, no one checking the rejections and no record.
Mistakes that repeat
Letting a tool reject people on its own. Using a score nobody can explain. Skipping the written criteria. Never reading what the tool turned down. Not telling candidates. Putting resumes into a personal account.
Frequently asked questions
Can AI screen resumes?
It can summarize and sort them against criteria you write. A person reads the rejections and makes every hiring decision.
Is it legal to use AI to screen resumes?
Often yes, but rules depend on where you hire. Some places require notice or an independent check. Ask your own adviser.
Should AI reject candidates automatically?
No. Have a person look at everyone the tool would pass on before a rejection is sent.
Can AI be unfair to candidates?
Yes. A tool can copy patterns from past hires or reward style over substance. The company is still responsible for the result.
Is it safe to put resumes into an AI tool?
Only an approved provider on a company account, with terms you have read, and a clear answer on how long the data is kept.
How do I start?
Write the criteria, ask for summaries instead of scores, read twenty resumes yourself to compare, and have a person check every rejection.
Is this legal advice?
No. Hiring rules differ by place. Ask your own adviser.
Not sure where you stand? Take the 3-minute AI readiness scorecard.