8 HR SOPs you can create with AI (and what to review)

AI can help an HR team turn scattered notes, policies, and process knowledge into a usable first draft. That is already a common use case: SHRM reported that 43% of organizations used AI for HR tasks in 2025, up from 26% in 2024. Among organizations using AI in recruiting, writing job descriptions was the most commonly reported application.

The useful word here is draft. An HR SOP shapes what managers do, what employees experience, and how the organization handles risk. AI can organize approved inputs and expose gaps, but it cannot know every policy, jurisdiction, exception, or unwritten dependency in your organization.

This guide covers eight HR SOPs you can create with AI, each with a copy-ready prompt and a practical review path. The prompts are designed to produce a working document, not a final policy or employment decision.

Before you prompt: Use only an AI tool your organization has approved. Remove personal, confidential, and regulated data. Confirm the governing source documents and name the human reviewer. If the work affects an individual employment decision or legal right, involve the appropriate HR, privacy, security, or legal expert before use.

Choose the review path before you draft

Risk depends on the content, the data you enter, how the output will be used, and which laws apply. Instead of treating an entire document type as universally “low risk,” assign the review path for the specific use case.

The prompts below all follow the same pattern: define the role, provide approved context, specify the task and format, set guardrails, and ask the model to identify missing information. In our AI for HR Professionals course, Jenelle Buatti teaches this as the CRAFT framework: context, role, action, format, and tone.

[[cta]]

8 HR SOPs you can build with AI

1. New-hire onboarding checklist

Onboarding is a strong place to begin because the process is repeatable and the source material usually exists. The challenge is getting steps out of inboxes, calendars, and people’s heads, then assigning each one to an owner.

Gather your current checklist, benefits deadlines, IT setup requirements, orientation schedule, and role-specific milestones. Sanitize the material before uploading it, and exclude information about an actual new hire.

Copy-ready prompt

You are an HR operations specialist helping us document a repeatable new-hire onboarding process.

Using only the approved source material below, create a phased onboarding checklist for [employee type or department]. Organize it into preboarding, Day 1, Week 1, First 30 days, First 60 days, and First 90 days.

For every task, include:
- task name
- purpose
- owner by role, not person
- due date or timing
- required input or dependency
- completion evidence

Do not invent policies, systems, deadlines, or owners. Put missing information in a separate “Questions for HR” section. Do not include employee names or other personal data.

Format the output as a table followed by the questions. Use a clear, supportive tone.

Approved source material:
[PASTE SANITIZED MATERIAL]

Review before use: Ask HR, IT, payroll, the hiring manager, and any other process owner to confirm their steps and timing. Then have someone unfamiliar with the process follow the checklist. That test will catch assumed knowledge that a read-through misses.

2. Operational employee handbook sections

AI can help draft operational handbook content when the underlying rules are already approved. Examples include where employees find technical support, how to request remote-work equipment, or how to use company-approved AI tools.

Do not ask a model to decide what your leave, accommodation, harassment, wage, discipline, or termination policy should say. Those areas may involve federal, state, local, contractual, and company-specific requirements.

Copy-ready prompt

You are an HR policy editor. Draft an employee handbook section about [TOPIC] using only the approved policy decisions and source documents below.

The section must include:
1. purpose and scope
2. who the section applies to
3. employee expectations
4. manager responsibilities
5. request or escalation process
6. related policies and document owner
7. review date placeholder

Do not add legal requirements, employee rights, disciplinary consequences, or exceptions that are not stated in the sources. If the sources conflict or leave a gap, flag it under “Questions requiring a decision.”

Write in plain language for employees at [COMPANY OR AUDIENCE]. Use a clear, respectful tone.

Approved source material:
[PASTE APPROVED, SANITIZED MATERIAL]

Review before use: The policy owner should compare every requirement with the source material. Route sections affecting employee rights, protected activity, pay, leave, accommodations, discipline, or separation through qualified counsel for the relevant jurisdictions.

3. Policy training quiz and knowledge check

Once a policy has been approved, AI can turn it into a short knowledge check. The model should work from the policy itself, cite the relevant section for each answer, and surface ambiguities instead of filling them in.

Copy-ready prompt

You are an HR learning designer. Using only the approved policy below, create a 10-question knowledge check for [AUDIENCE].

Include:
- five multiple-choice questions
- three scenario-based multiple-choice questions
- two true-or-false questions
- the correct answer and a one-sentence explanation
- the exact policy heading or section supporting each answer

Test practical understanding, not trivia. Do not introduce rules or examples that the policy does not support. If a question cannot be answered unambiguously from the policy, exclude it and list the ambiguity under “Policy gaps for review.”

Use accessible, neutral language.

Approved policy:
[PASTE APPROVED POLICY]

Review before use: The policy or compliance owner should verify every answer and source reference. Confirm that the quiz format, delivery, accessibility, recordkeeping, and required training content meet the organization’s actual obligations.

4. Job description and posting workflow

AI can turn an approved role profile into a consistent job description and posting. It can also flag vague phrases, unnecessary credentials, and requirements that do not clearly connect to the work. It should not decide on the qualifications or select candidates.

Federal employment protections apply across job advertising, recruitment, hiring, testing, compensation, performance evaluation, discipline, and termination. The EEOC’s overview of federal job-discrimination laws is a useful baseline, but employers must also account for applicable state and local laws.

Copy-ready prompt

You are a talent acquisition content specialist. Use the approved role information below to draft a job description and external job posting for [ROLE].

First, ask up to five clarifying questions about the role’s outcomes, essential functions, reporting line, work location, and required qualifications.

Then produce:
1. internal job description
2. external job posting
3. list of every qualification you included and its source
4. language audit flagging vague, exclusionary, gender-coded, age-coded, ableist, or unnecessarily inflated wording
5. questions HR or the hiring manager must resolve

Do not invent credentials, years of experience, physical requirements, compensation, benefits, or legal statements. Do not rank or evaluate candidates.

Use a specific, welcoming, plain-language tone.

Approved role information:
[PASTE APPROVED ROLE PROFILE]

Review before use: HR or talent acquisition and the hiring manager should confirm essential functions, minimum qualifications, compensation information, location requirements, and accessibility. Have the appropriate expert review any legal notices or jurisdiction-specific posting requirements.

5. Structured interview guide and scorecard

A shared interview guide helps interviewers assess the same job-related competencies against the same evidence. AI can draft questions and behavioral anchors from an approved competency model, but it should not determine who advances or generate questions about protected information.

The EEOC advises employers to train everyone involved in hiring and to avoid questions that can elicit prohibited information. AI-generated interview materials belong inside that governed process.

Copy-ready prompt

You are a structured-interview designer. Create an interview guide for [ROLE] using only the approved job description and competency definitions below.

For each of [NUMBER] competencies, provide:
- one behavioral question
- two job-related follow-up probes
- the evidence the interviewer should listen for
- behaviorally specific anchors for ratings 1, 3, and 5
- a note showing where the competency appears in the job description

Add a short interviewer introduction, instructions for consistent note-taking, and a final evidence summary. Flag any wording that could invite information about age, disability, medical history, family status, religion, national origin, or another protected characteristic.

Do not score a real candidate or infer personality, culture fit, health, or protected characteristics. List missing competency definitions separately.

Approved materials:
[PASTE APPROVED MATERIALS]

Review before use: HR or talent acquisition should validate that every question is job related and that rating anchors describe observable evidence. Train interviewers to use the guide consistently, make accommodations available, and document decisions according to policy.

6. Performance review process and templates

AI can build the container for a consistent review: the process steps, manager form, employee self-assessment, rating definitions, and calibration agenda. Managers and HR must still supply the evidence and exercise judgment about each employee.

Copy-ready prompt

You are an HR performance-management designer. Create a reusable performance review process and paired manager and employee forms using the approved competency model below.

Include:
1. process timeline and owners
2. employee self-assessment
3. manager assessment
4. rating scale with observable, role-neutral anchors
5. prompts for outcomes, behaviors, development, and support
6. calibration meeting agenda
7. completion and recordkeeping checklist

Do not write feedback about a real employee, infer intent, recommend a rating, or invent examples. Avoid personality labels and demographic assumptions. Put unresolved policy choices in “Decisions for HR.”

Use a direct, developmental, respectful tone.

Approved competency model and process rules:
[PASTE APPROVED MATERIALS]

Review before use: HR and representative managers should test whether the form works across roles and levels. Check for vague criteria, double standards, inaccessible language, and rating anchors that reward style or visibility instead of results and behavior.

7. Performance improvement plan framework

A model can help structure a blank PIP framework with sections for expectations, support, measurement, and check-ins. It should not decide whether a PIP is appropriate, draft allegations from an employee file, diagnose why performance changed, or recommend termination.

This is a high-review use case because it involves confidential information and may affect discipline or separation. Keep real employee data out of the drafting prompt unless your organization has expressly approved the tool and workflow for that data.

Copy-ready prompt

You are an HR process designer. Create a blank performance improvement plan framework for review by HR and employment counsel.

Include labeled sections for:
- purpose and applicable policy
- documented performance expectation
- objective evidence and dates
- specific improvement outcome
- measurement method
- support and resources
- check-in schedule
- review period
- employee response or acknowledgment
- possible next steps, marked “legal and HR language required”

Use placeholders only. Do not include a real employee, invent misconduct, draft a conclusion, recommend discipline, or supply legal language. Add reviewer fields for the manager, HR, and counsel, plus version and approval dates.

Format it as a reusable template followed by a completeness checklist. Use neutral, factual language.

Review before use: HR and qualified employment counsel should approve the framework and each application. Managers must use verified, job-related evidence; apply policy consistently; document support; and account for relevant accommodations, leave, agreements, and local requirements.

8. Offboarding checklist and exit interview guide

Offboarding combines routine coordination with legally sensitive obligations. AI can organize access removal, equipment return, knowledge transfer, and an exit interview guide. Payroll timing, benefits notices, records, severance, restrictive covenants, and separation agreements require the relevant experts.

Keep the operational checklist separate from exit-feedback analysis. If you analyze feedback, use deidentified, aggregated responses and follow the organization’s privacy, retention, and reporting rules.

Copy-ready prompt

You are an HR operations specialist. Using only the approved source material below, create two reusable documents for [EMPLOYEE TYPE OR LOCATION].

Document 1: Offboarding checklist
- organize tasks by notice received, final week, final day, and post-employment
- include owner by role, due date, dependency, completion evidence, and escalation path
- mark final pay, benefits, records, immigration, severance, and agreement-related items as “specialist validation required”

Document 2: Exit interview guide
- create eight open, neutral questions about role clarity, support, management, workload, growth, and culture
- include a voluntary-participation statement and note-taking guidance
- do not ask for medical, family, or other protected information

Do not invent deadlines, legal obligations, or company policy. List every missing input under “Questions for HR.”

Approved source material:
[PASTE APPROVED, SANITIZED MATERIAL]

Review before use: HR, payroll, IT, security, facilities, and the manager should verify the handoffs. Route jurisdiction-specific timing, notices, final pay, benefits, severance, and separation language to the responsible specialist or counsel.

Build a reusable HR SOP workflow

One good prompt can create a draft. A managed workflow makes the result repeatable.

1. Choose one process. Start with a recurring documentation bottleneck whose source materials and owner are clear.
2. Define the use boundary. State what AI may draft, what data is prohibited, who reviews it, and what requires escalation.
3. Build a source pack. Include only current, approved, sanitized policies, templates, terminology, and style guidance.
4. Set standing instructions. Tell the model to use only supplied sources, flag conflicts, expose missing information, and include document-control fields.
5. Draft and verify. Compare every requirement, deadline, citation, and system reference with its source.
6. Walk the process. Ask a person who performs the work to follow the SOP step by step.
7. Approve and publish. Record the owner, approvers, version, effective date, and next review date.
8. Measure and maintain. Track drafting time, review time, errors, completion rates, and employee questions. Review after a process or legal change, not only on an annual calendar.

If your approved AI platform offers a project or workspace feature, use it to store sanitized reference files and standing instructions. In the course, Maisha Cannon demonstrates this with a reusable “People Ops Content” project so new drafts begin with consistent context. Treat that workspace as governed documentation infrastructure, with a named owner and a process for removing outdated files.

What AI should not decide or publish alone

AI should not make final employment decisions or publish legal and compliance language without qualified review. That includes decisions and documents involving hiring, pay, accommodation, protected leave, harassment, investigations, discipline, performance action, layoffs, termination, severance, arbitration, and restrictive covenants.

The need for governance is not theoretical. In its 2026 survey, SHRM found that 57% of HR professionals working in states with employment-related AI rules were unaware of those rules. The same report found that legal and compliance teams most often led AI governance and oversight in respondents’ organizations.

Federal guidance points in the same direction. The U.S. Department of Labor’s AI principles for worker well-being emphasize transparency, meaningful worker engagement, worker-rights protections, and governance and oversight. The EEOC also explains how algorithmic tools can create disability-related risks in its AI and ADA resources.

Keep these boundaries in the workflow:

• AI may organize approved information; it does not establish legal requirements.
• AI may draft blank templates; it does not determine an employee outcome.
• AI may flag possible bias; it does not certify fairness or compliance.
• AI may summarize deidentified material; it does not receive unrestricted HR files.
• AI may propose questions; qualified humans verify facts, policy, law, and tone.

Protect employee and candidate data

“Enterprise” or “business” does not automatically mean “approved for every HR record.” Review the product, plan, settings, contract, retention controls, integrations, access rules, and your organization’s policies before entering confidential data.

Vendor terms also differ by offering and can change. For example, OpenAI says inputs and outputs from its business offerings are excluded from model training by default, while personal-workspace users have separate data controls. Anthropic says chats in its commercial offerings are not used for training unless the customer opts in. Those statements do not replace your own security, privacy, procurement, records, or legal review.

When in doubt, use placeholders and synthetic examples. “Employee A missed an approved performance target in two review periods” is safer drafting context than a name, compensation figure, medical detail, or pasted HRIS record. The finished document can be completed inside the organization’s approved system after review.

FAQs about HR SOPs and AI

Is it safe to create HR SOPs with AI?

It can be safe within a governed workflow. Use an organization-approved tool, provide sanitized and approved source material, prohibit unnecessary personal data, and assign a human reviewer. Security and privacy teams should validate the product and settings. Legally sensitive content and employment decisions require the appropriate HR or legal expert.

How long does it take to create an HR SOP with AI?

There is no defensible universal time estimate. Complexity, source quality, review depth, and legal exposure all matter. Measure your own baseline: capture drafting, verification, testing, and approval time before and after a pilot. A faster first draft is useful only when total cycle time and error rates also improve.

Which AI tool should HR use for SOPs?

Use a tool your organization has approved for the intended data and workflow. Evaluate the specific plan, training policy, retention, access controls, integrations, contract terms, and auditability. Even with a business product, minimize data and use placeholders unless privacy, security, legal, and records owners have approved the use case.

Can AI replace HR documentation work?

AI can reduce blank-page work, organize source material, standardize formatting, and flag gaps. HR professionals still own the facts, employee context, judgment, stakeholder review, compliance, and final approval. The strongest workflow assigns the model a narrow drafting role and makes human verification visible in the document’s control record.

Which HR SOPs need legal review?

Review needs depend on jurisdiction and use. Involve qualified counsel when an SOP addresses employee rights or employment decisions, including leave, accommodation, pay, harassment, investigations, discipline, termination, severance, arbitration, or restrictive covenants. Counsel may also be needed when automated tools influence recruiting, selection, promotion, performance, or workforce decisions.