HR AI literacy curriculum

A 4-module learning program for HR professionals at all experience levels. Designed to be delivered as a facilitated series over 4–6 weeks, or adapted for self-paced completion.

Total time: ~8 hours across all modules Format: Workshop + async reading + hands-on practice Audience: All People Team members Prerequisite: None

Program overview

Module Title Duration Format
1 How AI works, what HR professionals actually need to know 90 min Workshop
2 Using AI in your daily HR work 2 hours Workshop + lab
3 Governance, ethics, and responsible use 90 min Workshop
4 Building and scaling AI use cases 2 hours Workshop + project

Module 1: How AI works, what HR professionals actually need to know

Learning objectives:

Session outline (90 minutes):

Time Activity
0:00–0:10 Opening: What do you already believe about AI? (quick poll + discussion)
0:10–0:30 Concept: How LLMs work, without the jargon (facilitator presentation)
0:30–0:45 Concept: Hallucination, confidence, and why AI makes things up
0:45–1:05 Exercise: Spot the hallucination (participants identify errors in AI-generated HR content)
1:05–1:20 Discussion: Where should HR use AI? Where should we be cautious?
1:20–1:30 Wrap-up + homework: Try one AI tool before next session

Key concepts to cover:

What LLMs actually are: Large language models predict the next most likely word given context. They don’t “know” things the way a database knows things, they pattern-match against training data. This is why they can sound authoritative while being wrong.

The spectrum from rule-based to agentic:

Rule-based automation → AI Copilot → AI Agent
(deterministic)         (assisted)    (autonomous)

Most HR AI today sits in the middle. True agentic AI, where the model takes multi-step actions autonomously, is emerging and brings new governance requirements.

Prompt quality drives output quality: The single highest-leverage skill for HR professionals using AI is prompt construction. Vague prompts produce generic output. Specific, well-structured prompts with good context produce usable output.

Pre-reading (assign before session):

Module 2: Using AI in your daily HR work

Learning objectives:

Session outline (2 hours including lab):

Time Activity
0:00–0:15 Share-back from Module 1 homework, what did you try, what happened?
0:15–0:45 Prompt engineering for HR: the four elements of a good prompt
0:45–1:10 Live demo: Onboarding plan, performance review draft, policy Q&A
1:10–1:50 Hands-on lab: Each participant works through 2 prompts from the library using their own real work
1:50–2:00 Debrief: What surprised you? What didn’t work?

The four elements of an effective HR prompt:

  1. Role: Tell the AI who it is. “You are an experienced HRBP…” produces better output than just asking a question.

  2. Context: Provide the specific details the AI needs. Role, tenure, team size, situation. Don’t make the AI guess.

  3. Task: State clearly what you want it to produce. One output per prompt. “Draft a…” is clearer than “Help me with…”

  4. Constraints: Define length, tone, format, and what to avoid. “300 words, no bullet points, avoid generic phrases” significantly improves output quality.

Lab instructions:

Participants choose 2 prompts from the library that are relevant to their current work:

  1. Run the prompt with a real (or realistic) example from their work
  2. Evaluate the output: What’s useful? What’s wrong or missing? What would you change before using it?
  3. Iterate: Modify the prompt once based on what you learned and run it again
  4. Note: What changed between run 1 and run 2?

The verify-before-use rule:

Introduce and reinforce this heuristic:

Module 3: Governance, ethics, and responsible use

Learning objectives:

Session outline (90 minutes):

Time Activity
0:00–0:15 Case study: Three HR AI failures and what we can learn (discussion)
0:15–0:40 Framework: What makes AI use in HR risky? Legal and ethical dimensions
0:40–1:00 Scenario practice: Is this use case okay? (small groups, 4 scenarios)
1:00–1:20 Walk-through: The risk assessment template
1:20–1:30 Q&A + open discussion

Note on regulatory currency: Regulatory references in this module are current as of July 2026. The HR AI regulatory landscape is evolving rapidly, verify current law before relying on specific references in any compliance discussion or training material.

Key legal risk areas to cover:

Bias and disparate impact: AI trained on historical HR data can encode and amplify past discrimination. An AI that learned from 10 years of hiring decisions at a company where 80% of engineering hires were male will score resumes accordingly. Annual adverse impact testing is not optional, it’s legally and ethically required.

Illinois AI Video Interview Act, NYC Local Law 144, and emerging state laws: Several jurisdictions now require disclosure, impact assessments, or consent for AI use in hiring. Illinois’s 2026 amendment raised the bar: employers now need explicit written consent before AI analyzes a video interview, continuing the interview no longer counts as implied consent. Illinois also amended its Human Rights Act via HB 3773 (effective January 1, 2026) to broadly prohibit discriminatory AI use in hiring and promotion, not just video interviews. A companion law, Public Act 104-0425 (SB 2487, effective January 1, 2026), gives the Human Rights Commission tiered civil penalty authority for violations: up to $16,000 for a first offense, up to $42,500 with one prior violation within five years, up to $70,000 for two or more within seven years. The regulatory landscape is shifting quickly. Use cases involving hiring must be reviewed against current state law.

NYC Local Law 144 enforcement is entering a stricter phase. A December 2025 NY State Comptroller audit found DCWP’s enforcement “ineffective”: DCWP’s own review of 32 companies’ bias-audit disclosures caught 1 likely non-compliance issue, where the Comptroller’s review of the same 32 found at least 17. Expect more investigations and higher penalties as DCWP responds to the audit’s recommendations, not just the disclosure-and-audit paperwork the law describes on paper.

Colorado’s Automated Decision-Making Technology Act: Colorado SB26-189 (signed 14 May 2026, Colo. Sess. Laws ch. 131) repeals and replaces the state’s 2024 AI consumer protection law. It covers “consequential decisions,” including employment, made using automated decision-making technology (ADMT) that materially influences the outcome. Starting January 1, 2027, developers must give deployers technical documentation on intended use, training data categories, known limitations, and human review instructions. Deployers will owe consumers notice at the point of interaction, and within 30 days of any adverse consequential decision, a plain-language explanation plus the right to request human review and reconsideration. Records must be retained 3 years. The attorney general enforces under the Colorado Consumer Protection Act, with a 60-day cure period through January 1, 2030.

California: FEHA algorithmic discrimination rules and CPPA ADMT regulations: Two separate California regimes apply now, not one generic “CCPA” reference. The Civil Rights Council’s FEHA regulations on automated decision systems took effect October 1, 2025: they prohibit AI-driven disparate treatment or disparate impact in employment decisions, require four years of recordkeeping, and hold employers liable for discriminatory outcomes even when the tool comes from a third-party vendor. Separately, the CPPA’s ADMT regulations under the CCPA take effect January 1, 2027, and require pre-use notice, an opt-out right, and a formal risk assessment for AI used in hiring, promotion, or other significant employment decisions.

Texas: Responsible AI Governance Act (TRAIGA): Effective January 1, 2026 (HB 149, signed June 22, 2025). Narrower than Illinois or Colorado: TRAIGA prohibits intentional AI-based discrimination but does not create disparate impact liability, and the Attorney General has exclusive enforcement authority with a 60-day cure period before action.

GDPR and CCPA: Employees in covered jurisdictions have rights regarding automated decision-making. Know which of your employees are covered and how your AI use cases interact with their rights. See the California-specific rules above for the current state of play there.

UK: Data (Use and Access) Act 2025 replaces GDPR Article 22: Section 80 of the UK’s Data (Use and Access) Act 2025 took effect February 5, 2026 (Commencement Regulations SI 2026/82) and substitutes Article 22 of the UK GDPR with new Articles 22A to 22D. The new regime permits solely automated “significant decisions,” those with a legal or similarly significant effect, more broadly than before, but requires safeguards: notice to the individual, the ability to make representations, the right to human intervention, and the right to contest the decision. Stricter restrictions still apply where special category data is used. Treat this as a distinct regime from EU GDPR Article 22, not an equivalent, for any AI use case touching UK-based workers or candidates.

The ICO is already signaling how it will police this in recruitment specifically. A March 2026 ICO report on automated decision-making in hiring found many employers don’t recognize they’re doing ADM at all, and that “human review” in practice often amounts to rubber-stamping rather than a reviewer with real authority to change the outcome. Following that report, the ICO opened a consultation on draft ADM guidance, worth checking for the final version before relying on the March findings alone. Build the oversight role description to survive that scrutiny, not just to check the box.

Canada: no federal AI employment law, but Ontario requires disclosure: Canada’s federal AI bill (AIDA, part of Bill C-27) died when Parliament prorogued in January 2025 and has not been revived. AI regulation there runs through existing privacy, human rights, and sector law instead. Ontario is the exception: since January 1, 2026, employers with 25+ employees must disclose in publicly advertised job postings whether AI is used to screen, assess, or select candidates (Employment Standards Act, 2000 amendment).

The “AI as a manager” trap: AI that monitors, evaluates, or scores employees without their knowledge, even for legitimate purposes, creates legal and trust exposure. Transparency is not just ethical, it’s protective.

Scenario practice cards:

Print or share these four scenarios. Groups of 3–4 discuss: Is this use appropriate? What are the risks? What would you change?

Scenario A: An AI tool automatically rejects resumes with gaps of more than 6 months without showing them to a recruiter.

Scenario B: An HRBP uses an AI writing assistant to draft a PIP for an employee, then reviews and edits it before sharing with the manager.

Scenario C: People Analytics builds an attrition risk model that scores all employees monthly and shares the scores directly with managers.

Scenario D: An AI chatbot answers benefits questions for employees. It sometimes gives incorrect information but flags that users should confirm with HR.

Module 4: Building and scaling AI use cases

Learning objectives:

Session outline (2 hours):

Time Activity
0:00–0:15 Opening: What does “scaling AI in HR” actually mean?
0:15–0:45 Framework: From idea to deployment, the use case lifecycle
0:45–1:15 Team exercise: Prioritization matrix applied to your team’s top 5 ideas
1:15–1:45 Small group: Draft a one-page use case spec for your top-ranked idea
1:45–2:00 Gallery walk + feedback

Use case lifecycle:

Discover → Prioritize → Design → Pilot → Evaluate → Scale
   ↑                                          |
   └──────────────── iterate ─────────────────┘

One-page use case spec template:

Participants complete this for their team’s top-ranked use case:

Use case name:
Problem it solves (1 sentence):
Who benefits and how:
AI approach (Agent / Copilot / Automation / Analytics):
Data needed:
Human review gate:
Success metric(s):
What could go wrong:
Governance requirements (from risk assessment):
First step to move forward:

Assessment and certification

Participants who complete all four modules and the hands-on components receive a HR AI Practitioner acknowledgment from the People Team.

Optional extension: Participants who complete a full use case spec and present it to the HR Technology team are recognized as HR AI Champions an informal internal designation that supports peer enablement.

Facilitator notes