AI in Hiring and HR: Useful Tasks, Bias Risks, Legal Basics
AI can draft job ads, organize applications and answer routine HR questions in seconds. It can also quietly screen out good candidates and put your business on the wrong side of employment law. Here is how a small business can get the time savings without the risk.
Key takeaways
- Use AI freely for drafting and admin: job ads, interview guides, scheduling messages and policy summaries.
- Be very cautious with AI that ranks, scores or rejects people. That is where bias and legal exposure concentrate.
- Employers are generally responsible for discriminatory outcomes even when a vendor's software made the call.
- Rules differ by place. New York City, the EU and several US states have specific requirements, and they keep changing.
Hiring is one of the most time-consuming jobs for a small business owner. Writing the ad, sorting applications, scheduling interviews and answering employee questions can eat whole weeks. It is no surprise that recruiting platforms and HR software now include AI features, and that many owners already use general chat assistants for HR writing.
The trouble is that hiring decisions are also among the most regulated decisions a business makes. Anti-discrimination laws in most countries apply whether a person or a piece of software makes the choice. This guide separates the low-risk uses from the high-risk ones and gives you the legal basics to raise with an employment advisor. It is general information, not legal advice.
Low-risk tasks where AI saves real time
The safest uses are ones where AI produces a draft that a person reviews, and no individual candidate or employee is being judged by the tool.
- Job descriptions: turning your rough notes into a clear ad, then asking the tool to flag jargon, unnecessary requirements and wording that could discourage some applicants.
- Interview guides: generating structured, job-related questions and simple scoring rubrics so every candidate gets the same interview.
- Candidate communication: drafting acknowledgment emails, scheduling messages and polite rejection templates.
- Onboarding materials: first-week checklists, role summaries and training outlines.
- Policy drafts and summaries: turning a long handbook into plain-language FAQs, which a person then checks against the actual policy and local law.
- Employee survey analysis: grouping anonymous comments into themes, using the same careful process you would use for customer feedback.
Pro tip
Ask the AI to review your draft job ad as a skeptical candidate would. Prompt it to list requirements that seem unnecessary for the role, such as a degree for a job that does not need one. Removing them widens your applicant pool and reduces the chance of excluding people for reasons unrelated to the work.
High-risk tasks that need extra care
Risk rises sharply when a tool evaluates people. That includes:
- Resume screening that scores or filters applicants automatically
- Chatbots that ask knockout questions and reject candidates
- Video or game-based assessments that score personality or "fit"
- Tools that predict who will perform well, stay or leave
- Monitoring software that rates employee productivity
- Any system that recommends promotions, pay changes or terminations
These tools learn from data. If past hiring data reflects past bias, a model trained on it can repeat that bias at scale. A widely reported example is Amazon, which, according to a 2018 Reuters report, scrapped an experimental recruiting tool after finding it penalized resumes that included the word "women's." The company said the tool was never used on its own to evaluate candidates. The lesson for small businesses is simple: even well-funded teams found it hard to remove bias from automated screening.
Common mistake
Assuming that a vendor's "bias-free" claim moves legal responsibility to the vendor. In the US and many other places, the employer making the hiring decision is generally responsible for discriminatory outcomes, even when software designed by someone else produced them. Ask vendors for evidence, not slogans.
The legal basics to know
Laws on AI in employment are developing quickly, so treat this section as a starting map and check current rules where you hire.
United States: existing anti-discrimination law still applies
Federal laws such as Title VII of the Civil Rights Act and the Americans with Disabilities Act apply to hiring decisions however they are made. In 2023 the Equal Employment Opportunity Commission (EEOC) published technical assistance explaining how disparate impact rules could apply to AI selection tools, and earlier guidance on AI and disability. Law firms reported that the EEOC removed those AI documents from its website in January 2025 after a change in federal AI policy. The underlying laws did not change, however, so the core principle stands: if a tool screens out a protected group at a much higher rate and is not job-related, the employer can face liability.
New York City Local Law 144
New York City's Local Law 144, enforced since July 2023, covers "automated employment decision tools" used to substantially assist hiring or promotion decisions for jobs in the city. Employers must have an independent bias audit done within the year before use, publish a summary of the results, and notify candidates that such a tool will be used. A December 2025 audit by the New York State Comptroller criticized the city's enforcement as weak, which commentators expect could lead to tougher oversight.
Other US states
Several states have passed or proposed their own rules. Illinois has a law on AI analysis of video interviews, and other states, including Colorado and California, have adopted or proposed broader rules on automated decisions. Requirements and effective dates have shifted, so check the current status where your employees and applicants live.
European Union and UK
Under the EU AI Act, AI systems used for recruitment, selection and many employment decisions are classed as high-risk, which brings duties for both providers and the businesses deploying them. Separately, Article 22 of the GDPR gives people rights around decisions based solely on automated processing that significantly affect them, and job rejections can fall into that category. UK data protection law contains similar principles. Our explainer on AI regulations covers the EU timeline in more detail.
How to evaluate an AI hiring tool before you buy
If you are considering software that screens or scores candidates, ask the vendor these questions in writing:
- What exactly does it decide? Does it rank, recommend or automatically reject? Can a person override every outcome?
- What data was it trained on? Ask how the vendor checked that training data does not reflect historical bias.
- Has it been audited? Request recent bias audit results by sex, race and ethnicity where available, and ask who performed them.
- Can it explain its outputs? You should be able to tell a candidate in plain words what was assessed and why.
- How are accommodations handled? Candidates with disabilities must have a way to request an alternative format or process.
- What happens to candidate data? Check storage location, retention periods, training on your data and deletion on request.
- Which laws does it support? Ask whether the vendor helps with notices and audits required in places like New York City or the EU.
| Use case | Risk level | Minimum safeguard |
|---|---|---|
| Drafting job ads and templates | Low | Human edit and check for exclusionary wording |
| Interview question banks | Low | Keep questions job-related and identical for all candidates |
| Scheduling and reminders | Low | Offer a human contact for questions |
| Resume summaries for a recruiter | Medium | Human reads the original resume before any decision |
| Automated ranking or rejection | High | Bias audit, notice, human review and legal check |
| Employee monitoring or performance scoring | High | Transparency, local law review and clear limits |
Building a fair AI-assisted hiring process
A practical approach for a small business is to let AI handle the paperwork and keep people in charge of judgment. Write a structured scorecard for each role before you post it. Use the same questions and rating scale for every candidate. If you use AI to summarize applications, have the person making decisions read the original documents for anyone who is close to the line. Keep notes on why each finalist was chosen or rejected.
Never paste sensitive candidate information, such as health details, immigration status or background check results, into a general chat tool. Use approved HR systems for that data and check your tool's current privacy settings.
Hypothetical example
Imagine a 15-person bakery hiring a shift supervisor. The owner uses AI to draft the ad and a structured interview guide with five job-related questions and a 1 to 5 scale for each. The tool flags that "must lift 50 pounds" appears in the ad even though the role uses lifting equipment, so the owner removes it. All 40 applications are read by a person, and the five interviews follow the same guide. AI saved several hours of writing, and no software decided who got the job.
Benefits of careful AI use
- Faster, clearer job ads and candidate emails
- More consistent interviews with structured guides
- Less admin time for owners and managers
- Plain-language HR documents employees understand
Risks to manage
- Hidden bias in screening and scoring tools
- Legal duties for notices, audits and accommodations
- Sensitive personal data shared with the wrong tool
- Candidates put off by impersonal automated processes
Your AI hiring checklist
- List every hiring and HR task where AI is used today, including features inside existing software
- Mark each one as drafting, summarizing or deciding
- Keep a person in charge of every decision about an individual
- Ask vendors for bias audits, data terms and accommodation options in writing
- Check local rules where you hire, such as NYC Local Law 144 or EU requirements
- Tell candidates when AI is used and give them a human contact
- Review outcomes by group at least once a year
Frequently asked questions
Can I use ChatGPT or another chatbot to screen resumes?
You can technically paste resumes in, but it is risky. General chatbots are not designed or audited for hiring, outputs can vary, and you may be sharing personal data outside approved systems. If you do use AI for summaries, a person should read the originals and make the decision, and you should check the tool's current data terms.
Do I need a bias audit if I am a small company?
In New York City, Local Law 144 applies based on use of a covered tool for jobs in the city, not on company size. Other places have different thresholds. Even where no audit is required, checking selection rates by group is a sensible habit.
Do I have to tell candidates that I use AI?
Some laws require notice, including New York City's rule for covered tools and data protection laws such as the GDPR. Even where it is not required, being open about AI use builds trust and gives candidates a chance to ask for accommodations.
Is AI-written interview content legal to use?
Yes, as long as the questions are job-related and do not touch protected characteristics such as age, religion, disability or family plans. Review every AI-generated question yourself before using it.
Related reading
The bottom line
AI is a useful assistant for hiring paperwork and HR writing, and a risky judge of people. Use it to draft, organize and summarize. Keep humans responsible for decisions, ask vendors for proof of fairness, tell candidates when AI is involved and check the rules where you hire. That balance gives you the time savings without gambling your reputation or legal standing.
