Artificial intelligence is quickly becoming part of everyday hiring.
It can help employers draft job descriptions, organize applicant information, create interview questions, summarize candidate notes, schedule interviews, and communicate with applicants. Used thoughtfully, it can save time and help small businesses create a more consistent hiring process.
But AI is not a substitute for judgment.
Before using an AI tool to screen, rank, assess, or communicate with applicants, employers should ask five important questions.
1. What exactly is the tool doing?
“AI in hiring” can mean many different things.
Some tools simply help draft a job posting or organize interview questions. Others may scan resumes, rank applicants, score recorded interviews, recommend candidates, screen written responses, or automatically reject people based on selected criteria.
Those are very different uses.
Before turning on an AI feature, make sure you can clearly answer:
- Is the tool helping a human make a decision, or making a recommendation automatically?
- Does it rank, score, filter, or reject applicants?
- What information does it review?
- What factors affect an applicant’s score or recommendation?
- Can a human review the underlying information before a hiring decision is made?
If you cannot explain what the tool is doing in plain English, it probably should not be making—or heavily influencing—a hiring decision.
Federal employment-discrimination laws still apply when employers use AI and other algorithmic tools in hiring. The EEOC has specifically addressed the potential for these technologies to create discrimination concerns under laws including Title VII and the Americans with Disabilities Act.
The technology may be new. The employer’s responsibility is not.
2. Is the AI evaluating qualifications that are truly necessary for the job?
A hiring tool should focus on the actual requirements of the position.
That sounds simple, but automated tools can rely on signals that are not necessarily related to whether someone can succeed in the role.
A system might prioritize certain resume formats, specific word choices, uninterrupted work history, educational patterns, typing style, location, video-interview behavior, or other data points that have little to do with performing the job.
Before using AI to screen applicants, ask:
- Does this factor relate directly to the job?
- Is it a required qualification, a preferred qualification, or simply a nice-to-have?
- Could a qualified person be screened out because the tool is measuring the wrong thing?
- Does the tool consider transferable skills, not just a perfect keyword match?
- Would we be comfortable explaining this criterion to an applicant?
A seemingly neutral screening process can still create problems if it disproportionately excludes members of a protected group or relies on factors that are not sufficiently connected to the position.
There is also a more practical hiring problem: a system can be very efficient at eliminating exactly the kind of candidate a thoughtful recruiter might want to meet.
At Purciarele Group, we often talk about the value of looking beyond a perfect resume match. A candidate may not check every preferred box, but they may still have the transferable skills, judgment, work ethic, and potential to succeed.
Consider someone applying for a dental-office position who comes from a restaurant background. An automated system may see “no dental experience” and rank the person lower.
A human recruiter may see something different: years of client service, scheduling, problem-solving, reliability, communication, handling pressure, attention to detail, and professionalism.
That applicant may still need training. But they may also be an excellent hire.
AI should help employers identify qualified talent—not automatically eliminate people because they did not use the “right” words on a resume.
3. Who is reviewing the AI’s recommendation?
A good operating principle is simple:
AI can support the process. A person should own the decision.
Be cautious about allowing a tool to quietly reject applicants without meaningful human review, particularly for roles where transferable skills, communication, industry knowledge, personality, or trainability matter.
A human reviewer should be able to:
- Review the applicant’s actual resume, application, and relevant work history.
- Consider whether the candidate meets the true requirements of the role.
- Look beyond formatting, keywords, or automated rankings.
- Correct an inaccurate or incomplete AI recommendation.
- Make the final decision about who moves forward.
This does not mean every employer needs to abandon automation or manually redo every task an AI system performs.
It means someone should understand how the technology fits into the hiring process and remain accountable for the outcome.
The danger is not simply that AI might make a poor recommendation.
It is that an efficient system can make the wrong decision very efficiently.
The goal is not to remove people from hiring. It is to help people make better, more efficient decisions.
4. Could the tool create an accessibility or discrimination concern?
AI can create risk when it relies on characteristics that do not actually measure whether an applicant can perform the job.
That concern deserves particular attention when a tool evaluates or interprets things such as:
- Recorded video interviews.
- Facial expressions or eye contact.
- Voice, speech, tone, or accent.
- Typing speed.
- Online assessments.
- Personality or behavioral assessments.
- Gaps in employment history.
- Automated resume screening.
- An applicant’s interaction with technology or assistive technology.
For example, an automated assessment could disadvantage an applicant with a disability even though that person could perform the essential functions of the job with or without a reasonable accommodation.
Employers should understand how applicants can request an accommodation if an automated assessment, timed test, video interview, or other part of the hiring process creates a disability-related barrier.
A clear statement can help:
If you need a reasonable accommodation to participate in the application or interview process, please contact [name/email/phone].
Of course, putting the sentence on an application is only the first step.
Someone also needs to be responsible for receiving those requests, responding to them, and making sure the accommodation process actually works.
5. Can you explain and defend the process?
Before using AI in hiring, ask yourself:
If an applicant asked how we made this decision, could we explain our process professionally and truthfully?
You do not need to reveal proprietary technology or debate every hiring decision. But you should be able to explain, at a high level:
- The job-related qualifications the company considered.
- Whether AI or automation was used to support part of the process.
- How human review fits into the process.
- How applicants can request a reasonable accommodation.
- Who is accountable for the final hiring decision.
This is also a vendor-management question.
Before purchasing or activating an AI screening tool, ask the vendor:
- What information does the tool use to score, rank, or filter applicants?
- Can we turn off automatic rejection?
- Can we review applicant data and override recommendations?
- Has the tool been evaluated for bias or adverse impact?
- How does the tool support reasonable accommodations?
- Where is applicant data stored, and who can access it?
- Does the vendor use applicant information to train its own AI models?
- What retention and deletion options are available?
- Can the vendor provide documentation about the tool’s purpose, limitations, testing, and validation?
A polished demonstration is not the same thing as understanding how a product works.
Employers should know enough about the tool to decide whether it belongs in their hiring process in the first place.
The National Institute of Standards and Technology’s voluntary AI Risk Management Framework offers a useful way to think about these questions. It encourages organizations to consider issues including validity and reliability, accountability and transparency, privacy and security, explainability, and fairness when selecting and using AI systems.
One more consideration: location matters
Federal law is not the whole picture.
State and local requirements may also affect how employers can use AI or automated employment tools. Depending on the jurisdiction, employers may face additional requirements involving notices, assessments, documentation, privacy, or other aspects of the hiring process.
Before implementing a new hiring technology, employers should determine what requirements apply based on where they operate and where they hire.
AI should support good hiring—not replace it
AI can be useful.
It can help employers save time, create consistency, organize information, and reduce repetitive administrative work.
But the technology does not remove the employer’s responsibility.
The employer still owns the job requirements.
The employer still owns the hiring process.
The employer still owns the final decision.
And the employer is still responsible for making sure applicants are evaluated fairly, consistently, and based on qualifications that actually matter for the role.
Before using AI in hiring, remember these five questions:
- What exactly is the tool doing?
- Is it evaluating qualifications that are truly necessary for the job?
- Who is reviewing the AI’s recommendation?
- Could the tool create an accessibility or discrimination concern?
- Can we explain and defend the process?
AI can make hiring more efficient.
Human judgment keeps it thoughtful.
We Love HR So You Don’t Have To®
