Most guides to AI in HR list a dozen use cases and a few vendors, then mention bias in a single paragraph at the end. That order is backward. The use case you pick decides your legal exposure: an AI that drafts a benefits FAQ carries almost no risk, while one that ranks job applicants can trigger bias audits, candidate notices and lawsuits.
This guide puts the two together. You get a plain explanation of artificial intelligence in HR, the 2026 adoption numbers, the main use cases with sourced examples, the tools by category, a table that maps each use to its risk, a current summary of the AI hiring laws (as of November 2026), and a checklist for buying and rolling out a tool.
What is artificial intelligence in HR?
Artificial intelligence in HR is the use of software that finds patterns in data or generates language to support human resources work. It covers three kinds of tools. Predictive machine learning scores or ranks things, such as which applicants match a role or which teams show turnover risk. Generative AI writes and summarizes, as explained in our guide to the definition of generative AI. AI agents take actions inside HR systems, such as filing a time-off request or generating a letter.
The distinction matters because the risk rises as the AI moves from helping a person write to helping a person decide. A drafting tool's mistakes are visible on the page. A ranking model's mistakes are hidden inside a shortlist that nobody sees in full.
AI in human resources by the numbers (2026)
SHRM's State of AI in HR 2026 (opens in a new tab) research, based on 1,908 HR professionals and published in March 2026, is the most current broad survey. Its headline finding is that HR is behind the rest of the business.
Just 39% of organizations have implemented AI in their HR functions, with another 7% planning to do so this year.
- 62% of organizations use AI somewhere in the business, and 31% have no plans for AI in HR.
- Where HR uses AI, it is concentrated in recruiting (27%), HR technology (21%) and learning and development (17%).
- 56% of HR functions do not formally measure the success of their AI investments, and only 16% use ROI as a metric.
- 87% of HR professionals report improved efficiency from AI, and 75% report better work quality.
- 57% of HR professionals in states with workforce AI regulations are not aware of those rules, and only 12% of those who are aware have taken steps to comply.
Those last two numbers are the gap this guide is meant to close: HR teams report real efficiency gains, but most are not measuring results or tracking the rules that apply to them.
How to use AI in HR: the main use cases
Recruiting and talent acquisition
Recruiting is where AI in HR started and where it is most used. Common jobs include writing and checking job ads, sourcing candidates, screening applications against requirements, scheduling interviews and answering candidate questions. LinkedIn's Hiring Assistant (opens in a new tab), which it calls its first AI agent for recruiters, became available globally in English in September 2025. LinkedIn says early adopters were saving more than 4 hours per role and reviewing 62% fewer profiles. Those are the vendor's own figures, and the recruiter still decides who moves forward.
Screening and ranking is also the riskiest HR use of AI. Amazon learned this early: according to MIT Technology Review (opens in a new tab), the company began building a system in 2014 to rate applicants from one to five stars, trained it on 10 years of résumés, and scrapped it after finding it penalized résumés containing the word "women's." A model trained on past hiring decisions learns past hiring patterns, including the ones you would never write into a policy.
HR help desks and employee self-service
Employees ask the same questions about leave, pay and benefits thousands of times a year, which makes the HR help desk a strong fit for AI. IBM's internal AskHR (opens in a new tab) is the best-documented example. IBM reports a 94% containment rate on common questions, more than 11.5 million employee interactions in 2024, around 80 automated HR tasks, and a 40% reduction in the HR team's operational costs over four years. These are IBM's figures about its own product. The design lesson is the useful part: routine requests go to the AI, and complex cases go to human advisors. Our roundup of agentic AI examples covers how AskHR routes requests.
Onboarding and employee experience
AI can assemble onboarding checklists, answer new-hire questions from the handbook, and trigger account and equipment requests across systems. The safest approach is to let the AI work from your approved documents only, a pattern called retrieval-augmented generation, so answers come from your policy rather than the model's guesses.
Generative AI in HR: drafting and summarizing
This is the lowest-risk place to start. HR teams use general assistants to draft job descriptions, interview guides, policy summaries, offer letters and manager talking points, and to summarize engagement survey comments. The output is a draft a person edits, so errors are visible. A structured prompt gives better, more consistent results than a one-line request. Here is an example you can adapt:
Role: Senior Payroll Specialist, hybrid, reports to Payroll Manager.
Must-have skills: multi-state US payroll, year-end reporting, our HRIS.
Nice-to-have: process automation experience.
Write a job description under 400 words with sections for
About the role, What you will do, What you bring, and Benefits.
Rules:
- Use plain, gender-neutral language.
- List only requirements I gave you. Do not add degrees or years of experience.
- Flag any phrase that could discourage qualified applicants from applying.Workforce analytics and performance
AI can answer plain-language questions about headcount, turnover and time off, and flag patterns such as rising attrition in one team. When the same tools start scoring individual employees for promotion, discipline or termination, the risk profile changes completely, because those are employment decisions under most of the laws below.
AI tools for HR by category
You rarely need a new platform to start. The fastest route is usually the AI features already inside your HR system. The table lists examples verified on vendor pages in November 2026; it is a map of categories, not a ranking. For agent platforms outside HR, see our guide to the best AI agents.
| Category | Example | What the vendor says it does |
|---|---|---|
| HR suite with AI agents | Workday (opens in a new tab) | Agents for HR, including Talent Acquisition, Payroll and Workforce Management agents |
| HR suite assistant | SAP Joule (opens in a new tab) | Handles everyday HR tasks such as requesting and approving time off and giving feedback |
| Small and midsize HRIS | BambooHR (opens in a new tab) | Plain-language questions and reports, plus flags for burnout and turnover risk |
| Recruiting agent | LinkedIn Hiring Assistant (opens in a new tab) | Sources candidates for recruiters and learns from their feedback |
| HR help desk agent | IBM AskHR (opens in a new tab) | IBM's internal agent that answers HR questions and completes HR tasks |
| General assistant | Any business-grade LLM assistant | Drafting, summarizing and analysis on documents you provide |
Two cautions. First, vendor performance numbers are self-reported; ask for results from customers like you. Second, an AI agent that can change records in your HR system needs the same access controls as a new employee with admin rights. Our guide to agentic AI security covers permissions and prompt injection.
Match each HR use case to its risk
This is the table most AI in HR guides leave out. It shows how far each use is from an employment decision, and what that means in practice.
| Use case | Risk level | Why | Minimum safeguard |
|---|---|---|---|
| Drafting job ads, policies, letters | Low | A person reviews and edits every output | Human edit; check wording for bias |
| HR help desk answers | Low to medium | Wrong answers about pay or leave cause real harm | Answer only from approved documents; easy handoff to a person |
| Onboarding workflows | Low to medium | Touches personal data and system access | Limit permissions; log actions |
| Interview scheduling and candidate Q&A | Medium | Candidate-facing, but does not judge candidates | Tell candidates they are talking to AI |
| Sourcing and targeted job ads | High | Listed as high-risk in the EU AI Act | Check reach across groups; document criteria |
| Screening, ranking or scoring applicants | High | Covered by NYC, Illinois, Colorado and EU rules | Bias audit, notices, human review of rejections |
| Promotion, discipline or termination support | High | Employment decisions under every law below | Human decision-maker; explanation on request |
| Emotion recognition of employees | Prohibited in the EU | Banned in EU workplaces since February 2025 | Do not use, except medical or safety cases |
AI hiring laws to check before you deploy (as of November 2026)
Rules are spreading and changing quickly. SHRM notes that 19 of the most populous US states have enacted AI laws or regulations relating to employment. These are the ones most often cited. This is general information, not legal advice; check with employment counsel for your locations.
| Jurisdiction | Rule | Status | What it requires |
|---|---|---|---|
| New York City | Local Law 144 | Enforced since July 5, 2023 | Bias audit within one year before use, public audit summary, notice to candidates 10 business days before use |
| Illinois | HB 3773 (Human Rights Act amendment) | In effect since January 1, 2026 | No AI use with a discriminatory effect, no zip codes as a proxy for protected classes, notice to employees when AI is used |
| Colorado | SB 26-189 (replaced SB 24-205) | Takes effect January 1, 2027 | Pre-use notice, explanation of adverse decisions, human review where commercially reasonable, three years of records |
| European Union | AI Act, Annex III point 4 | High-risk duties apply from December 2, 2027 | Risk management, data governance, human oversight and transparency for hiring and worker-management AI |
| European Union | AI Act, Article 5 | In effect since February 2, 2025 | Ban on AI that infers emotions in the workplace, except for medical or safety reasons |
New York City Local Law 144
The city's Department of Consumer and Worker Protection (opens in a new tab) says employers may not use an automated employment decision tool unless it has had a bias audit within one year of use, the audit information is public, and candidates or employees have received notice. Enforcement began on July 5, 2023. A December 2025 audit by the New York State Comptroller called that enforcement "ineffective." As summarized by DLA Piper (opens in a new tab), 75% of test calls to the 311 hotline about AI hiring tools never reached the agency, and the Comptroller found at least 17 potential compliance issues in 32 audits where the agency had found one. The agency agreed to most of the recommendations, so expect enforcement to tighten.
Illinois and Colorado
Illinois amended its Human Rights Act, effective January 1, 2026. According to Morgan Lewis (opens in a new tab), employers may not use AI that has the effect of discriminating against protected classes, may not use zip codes as a proxy for them, and must notify employees when AI is used in decisions from recruitment and hiring to promotion, discipline and discharge. The Illinois Department of Human Rights sets the details of the notice.
Colorado's original AI Act never took effect. On May 14, 2026, the governor signed SB 26-189, which repeals and replaces it from January 1, 2027. As Arnold & Porter (opens in a new tab) explains, employers no longer need a detailed risk management policy or impact assessments, but they must give notice before AI materially influences an employment decision, explain adverse outcomes, and offer human review where commercially reasonable. The attorney general enforces it, with penalties of up to $20,000 per violation. In July 2026 the FTC proposed a policy statement suggesting the law may be preempted, which leaves its future uncertain.
The EU AI Act
The EU AI Act lists employment as a high-risk area. Annex III (opens in a new tab) covers AI used for recruitment or selection, including targeted job ads, filtering applications and evaluating candidates, and AI used for decisions on promotion, termination, task allocation or monitoring performance. These obligations were due on August 2, 2026, but the Digital Omnibus on AI, Regulation (EU) 2026/1744, moved them to December 2, 2027 (opens in a new tab). The ban on emotion recognition in the workplace under Article 5 (opens in a new tab) has applied since February 2, 2025.
Existing discrimination law still applies
You do not need an AI-specific law to be liable. In Mobley v. Workday, in federal court in Northern California, an applicant alleges that AI screening tools unlawfully screened out qualified applicants, including under the federal Age Discrimination in Employment Act. In early 2026, Forbes reported (opens in a new tab) that the court had authorized notice to potential opt-in plaintiffs. Workday says its AI does not make hiring decisions and is not designed to reject candidates automatically. The claims are unproven, but the case shows that vendors and employers alike can face scrutiny for automated screening.
A checklist for buying an AI tool for HR
- Name the task and the metric. Write down the one job the tool will do and how you will measure it, such as time to fill, tickets resolved without escalation or hours saved per week. SHRM found most HR teams skip this step.
- Classify the risk. Use the table above. If the tool influences hiring, promotion or termination, treat it as high risk from day one.
- Ask for bias testing evidence. Request the vendor's most recent bias audit or adverse impact analysis, the data it was based on, and whether it meets NYC Local Law 144 if you hire in the city.
- Check notices and explanations. Confirm the tool lets you notify candidates and employees and explain how it influenced a decision, as Illinois and Colorado require.
- Keep a human in the decision. Configure the tool so a person reviews rejections and every promotion or termination recommendation.
- Review data handling. Ask where employee data is stored, whether it is used to train the vendor's models, and how long it is kept.
- Limit agent permissions. Give AI agents the least access they need, and log every action they take in HR systems.
- Pilot, measure, then expand. Run one team or one requisition type for a set period, compare results with your baseline, and only then roll out more widely.
If you are building agents rather than buying them, our guide to AI agentic workflows shows how to keep a fixed process with AI steps inside clear limits, which suits most HR processes. For the wider business case, see AI agents for business.
The bottom line
Artificial intelligence in HR pays off first on drafting, help-desk questions and routine requests, where a person can easily check the output. Hiring and promotion uses can save time too, but they come with bias audits, notices and human review in a growing number of places, and existing discrimination law applies everywhere. Start low-risk, measure what changes, and move toward decision support only when you can show it works fairly. For more practical guides on AI tools and agents, browse our AI guides.