AI hiring for a small business usually starts without a decision. The founder asks a chatbot for a job post, the applicant tracking system switches on an AI ranking feature, and nobody on a 12-person team is checking whether any of it creates legal exposure.
Key Takeaways
- AI hiring for a small business with no HR person is safe for drafting job posts, scheduling interviews and summarizing interview notes, and needs a named human on every screening and rejection decision.
- The legal exposure comes from AI that ranks, scores or filters candidates, because that is the step anti-discrimination law looks at, whatever the size of the firm.
- As of September 2026, federal anti-discrimination law under Title VII applies to employers with 15 or more employees, while many state laws reach smaller firms, and New York City’s Local Law 144 turns on where the job is located.
- A founder can check their own exposure with 3 questions: how many employees they have, where the role is based, and whether any tool in the process scores or rejects applicants without a person reading them.
- A one-page hiring record that shows who reviewed each rejection, and on what criteria, is the cheapest protection a firm without HR can keep.
Which recruiting tasks AI can take in a small business
AI can take the writing and coordination work around hiring. It should not decide who moves forward. That line is simple enough for a founder to hold without an HR person, and it matches where both the time savings and the legal risk sit.
The table sorts the common recruiting tasks by how much of each can go to AI.
| Recruiting task | AI role | Human role | Legal sensitivity |
|---|---|---|---|
| Writing the job post | Drafts from your role notes | Removes requirements you do not need | Low, check for age or gender-coded wording |
| Posting and sourcing | Formats for each job board | Picks the boards | Low |
| Reading applications | Summarizes each one against your criteria | Reads every application it summarizes | High if AI ranks or rejects |
| Shortlisting | None, or a summary table only | Decides the shortlist | High |
| Scheduling interviews | Books slots, sends reminders | None | Low |
| Interview notes | Transcribes and summarizes with consent | Writes the hiring decision and reason | Medium, recording consent rules vary by state |
| Rejection emails | Drafts a standard message | Approves each send | Medium |
| Reference checks | Drafts questions | Makes the calls | Low |
The pattern across the table is that AI prepares material and a person makes each call about a candidate. The AI for HR and people operations guide covers the wider people function, including onboarding and performance. This post stays with the hiring decision.
What AI does well in hiring
AI saves the most time on the steps founders put off. A job post that sits in a draft for 2 weeks costs 2 weeks of applicants.
Job posts
Give the model your notes on the role, the actual tasks in a normal week, and the pay range. Ask it to write the post and then list every requirement it included. Delete any requirement you cannot connect to a real task. Degree requirements and years-of-experience minimums tend to creep in from the model’s training data, and they shrink your pool for no reason.
Scheduling
Scheduling is pure coordination. A booking link tied to your calendar, plus automatic reminders, removes the email chain completely. This is the safest place to start.
Interview notes
Transcription tools that summarize an interview give a founder a record they would otherwise never write. Tell the candidate at the start that the call is recorded, and check the consent rules in your state and theirs, because some states require every party to agree. Use the summary as a memory aid. Write your own 3-line decision note separately.
Where AI hiring creates legal exposure for a small business
The exposure starts when software ranks, scores or filters people. A resume screener that pushes some applicants to the bottom of a list is making a selection decision, and selection decisions are what discrimination law covers.
The risk is disparate impact: a neutral-looking rule that screens out one group more than others. An AI model trained on past hiring data can learn that pattern without anyone seeing it. The employer carries the liability, and a vendor’s assurance that the tool is fair does not move it. AI vendor and model risk covers the questions to ask a vendor before you switch a feature like this on.
Many applicant tracking systems now include AI ranking as a default setting. Check yours. If it sorts applicants by a match score, either turn the score off or make sure a person reads every application, including the low-scored ones.
Which hiring rules apply at your size and location
Three facts decide most of it: headcount, where the role is based, and what the tool does. This section is a starting point for a conversation with an employment lawyer and is not legal advice.
Headcount. Title VII of the Civil Rights Act, which the EEOC enforces, applies to employers with 15 or more employees. The age discrimination law starts at 20. Many state laws start lower, some at 1 employee, so a 6-person firm is rarely outside every rule.
Location of the role. New York City’s Local Law 144 covers automated tools that substantially assist hiring decisions for roles based in the city, including remote roles tied to a city office. It requires an independent bias audit within the past year, a published summary of the results, and advance notice to candidates. The law has no small-employer exemption. The city’s Department of Consumer and Worker Protection publishes the rule text, and that primary page is the one to read. Illinois also amended its Human Rights Act to cover AI in employment decisions from January 2026.
What the tool does. A tool that drafts a job post or books a meeting is outside these rules. A tool that scores, ranks or rejects applicants is inside them.
| Your situation | Likely exposure | Question for counsel |
|---|---|---|
| Under 15 staff, no AI screening | Low | Which state rules apply at my size? |
| Under 15 staff, AI ranks applicants | State law, possibly city law | Does my state cover AI-driven selection? |
| 15 or more staff, AI ranks applicants | Federal and state | Can I show a person reviewed every rejection? |
| Any size, hiring for a NYC-based role with an AI screener | Local Law 144 | Do I need a bias audit before using this tool? |
Federal guidance on AI in hiring has changed since 2025, and some earlier EEOC technical guidance is no longer published. The underlying law has not changed, and a claim of discrimination is judged against it.
How to keep a human in the hiring decision when you are the only reviewer
Read every application yourself, or have one named person read them, and record why each person moved forward or did not. At a firm with 20 applicants for a role, that is about 2 hours. The record matters more than the time.
A founder-sized version:
- Write 4 to 6 criteria for the role before the post goes live. Tie each one to a task in the job.
- Let AI produce a summary of each application against those criteria.
- Read the application itself, then mark each criterion met or not met.
- Write 1 line for each rejection that names the criterion it missed.
- Keep the sheet with the role for at least a year.
This is the same checkpoint covered in human-in-the-loop AI review, applied to the decision where a missed check costs the most. If the team uses AI tools in hiring on their own, add a line to your AI acceptable use policy saying which ones are approved.
Setting up AI hiring in a week
A founder can put this in place in 5 working days before the next role opens.
- Day 1: check your applicant tracking system for any AI ranking or auto-reject setting, and turn off anything that rejects without a person.
- Day 2: connect a booking link and reminders for interviews.
- Day 3: write the criteria template and the 1-line rejection record.
- Day 4: draft a job post with AI and cut every requirement you cannot tie to a task.
- Day 5: send your size, your hiring locations and your tool list to an employment lawyer and ask the questions in the table above.
If the hire is being made to cover routine work, check first whether the work needs a person at all. Scaling a business with AI instead of hiring covers that decision. For how hiring connects to capacity, payroll and the rest of the back office, see the AI-powered back office design guide.
Frequently Asked Questions
Is it legal for a small business to use AI to screen resumes?
Yes, with care. Using AI to summarize applications is low risk. Letting it rank or reject applicants brings the tool under anti-discrimination law, and in New York City it can trigger a bias audit requirement, so keep a person on every rejection.
Does NYC Local Law 144 apply to my business if I am not based in New York?
It can. The law turns on where the role is located, so a remote role tied to a New York City office can be covered even if the company is based elsewhere. Read the city’s rule text and confirm with counsel before using an AI screener for that role.
Am I too small for AI hiring rules to apply to me?
Probably not for every rule. Federal Title VII starts at 15 employees, but many state anti-discrimination laws cover much smaller employers. City rules like Local Law 144 have no size exemption at all.
Can AI write my job posts?
Yes, and it is one of the safest uses in hiring. Give it the real weekly tasks and the pay range, then delete any requirement you cannot connect to a task. Check the wording for phrases that signal age or gender preferences.
Do I need to tell candidates I use AI in hiring?
In some places, yes. New York City requires advance notice when an automated tool is used in the decision, and recording an interview needs consent in many states. Telling candidates is a sound default everywhere.
