AI for HR
Build two working AI assistants for your HR team with your own hands — no coding, no cost, and nothing that stays in the classroom.
A hands-on, no-code course that takes HR professionals from AI-curious to AI-capable in two sessions. You build a personal AI Recruiter that screens and ranks candidates for one recruiter — and learn why one flat knowledge base silently fails, and how purpose-built ones fix it — then a team HR Generalist that answers policy questions, looks up records, and files tickets for the whole HR function, with a guardrail on every capability. You measure your agent with a balanced eval set, and you learn the EU AI Act and GDPR rules that make every guardrail a legal requirement. Everything is grounded in a complete fictional company corpus, and everything used is free and runs in your browser.
Who it's for — HR professionals, HRBPs, talent acquisition partners, and HR leaders who want to use AI to do their actual job better — no coding background required.
In class you work on a synthetic company corpus — a fictional business modeled on real-world challenges, so you practice on realistic, messy data, not toy examples.
- Explain how AI models work, why they hallucinate, and what HR must do about it — in your own words, to your CEO or board
- Put the three data-privacy questions to any AI vendor — training, storage, accountability — and pick the tier allowed to touch HR data
- Map your HR processes to where retrieval, agents, and human review actually belong — and where they don't
- Build an AI Recruiter end-to-end in Dify: draft a JD against real skill gaps, screen and rank real résumés with cited evidence, cross-check interview minutes
- Structure knowledge so agents answer completely: see a flat knowledge base silently drop candidates, then fix it with purpose-built KBs
- Build a team HR Generalist: cited policy answers, gated record lookups, and tickets that never file without human confirmation
- Run a balanced eval set against your agent — and keep running it on every prompt, policy, or model change
- Apply EU AI Act and GDPR rules to AI in HR — know what's high-risk, what triggers Article 22, and your deployer obligations
You leave with Two working HR assistants — a personal recruiter for one role-owner and a team generalist for the whole HR function.
These aren't demos — they're working agents you build hands-on and tailor to your own role, data, and custom spec.
AI Recruiter
A personal screening agent that drafts the job description from your actual skill gaps, ranks candidates with cited evidence, and flags discrepancies between what candidates claim and what the interviews show — always pausing for human review before anyone is contacted.
HR Generalist
A team agent that answers policy questions with citations, looks up records with access gated by who is asking, and drafts HR tickets that only file once a human approves — it routes, it never approves.
How you build them, class by class
Foundations, and Your First AI Assistant for HR
- How LLMs generate text, why they hallucinate — and the data-privacy questions to ask before HR data goes anywhere
- Prompt design as a guardrail: build one job-description prompt from weak to strong, technique by technique
- Be the retrieval engine yourself first, then build a knowledge base over a realistic HR corpus and prove it answers with citations
- Build the AI Recruiter: draft the JD, screen and rank candidates with evidence, cross-check claims against interview minutes
- Watch a flat knowledge base fail honestly — a confident, incomplete shortlist — and rebuild on purpose-built KBs so ranking works
Build the HR Generalist, Then Measure and Govern It
- One personal agent, one team agent: the Generalist serves everyone, with access decided by who is asking
- Build the Generalist on its guidebook: cited policy answers, record lookups that apply the policy on top, and tickets routed to a human — never auto-approved
- Live data over MCP: RAG is memory, MCP is reach — and the four failure modes every system prompt must cover
- Run a balanced ten-question eval — coverage, diversity, criticality — and treat it as a standing check, not a launch gate
- EU AI Act and GDPR: risk tiers and the timeline, deployer obligations, the red lines, the penalties — and the frontier debate on identity and logging
The names the whole industry runs on.
Dify
One of the most popular open-source platforms for building AI agents — you assemble your recruiter and generalist visually, no code required.
Gemini 2.5 Flash
Google's fast, low-cost model and the backbone of every assistant you build — its free tier covers the entire course.
MCP (Model Context Protocol)
The fast-emerging open standard for connecting AI to live data and actions — how an agent fetches a current balance instead of guessing from a stale file.
Plus the rest of the working stack — Dify · Gemini 2.5 Flash · text-embedding-004 · Claude · ChatGPT · Gemini · MCP (Model Context Protocol) · A corpus-driven eval-set generator.
Your hands-on capstone: two working AI assistants — the personal AI Recruiter and the team HR Generalist — plus a graded take-home screening of a full requisition and an eval run proving how good your agent actually is. The work is grounded in a realistic company corpus throughout. This is what your certificate is based on — not attendance.
Your certificate is based on what you build — not attendance.
Ready to build your AI agent?
A small, high-quality cohort for professionals — fully online, hands-on from minute one.