AI Essentials for Human Resources

1 x 2.5 Hour Bespoke Session (online, with one 15 minute break)

Built for HR teams and stands on its own, with seven use cases run end to end across recruitment, staff queries, communication and development.

Summary:

A bespoke session for one of the most data-rich functions in any organisation, built around the questions an HR team actually asks: can AI screen CVs, can it stop our interview questions repeating themselves, can it send our candidate letters, and what would it recommend for someone’s development. It runs on the two things generative AI is genuinely good at, recognising patterns and personalising at scale, and on the three pillars of any chatbot: instructions, a knowledge base and tools. Ethics and hallucination are dealt with early and honestly, with real cases of AI recruitment going wrong, and the session ends on the hardest question of the lot, screening, with a properly argued answer rather than a comfortable one.

Course Breakdown:

Foundations, and the careful part

• HR as a data-rich function: recruitment and development
• HR as a data-rich function: staff queries, reporting and everyday admin
• Introduction to generative AI
• Hallucination: why AI makes mistakes, and how to make it quote its source and flag what is missing
• Ethics: you are handling people, not documents, covering bias, transparency and over-reliance, with real cases
• Prompt engineering, and the four parts of an effective prompt: what, who for, based on what, and in what shape
• When it is worth reaching for AI at all, and the two directions in which it helps

Pattern recognition

• Use case: writing job descriptions, dictated messily against a previous notice used as the shape, with an assumptions list to check before it goes out
• The three pillars of a chatbot
• What a knowledge base is
• Use case: generating interview questions from the description just written, grouped by criterion with a scoring guide, then anchored to a library of previous sets so they stop repeating themselves

Personalisation

• Custom instructions versus memory, and when to use which
• Use case: recruitment posts, one vacancy written four ways for four different readers
• What a connector is
• Use case: personalised email management, letter templates loaded once so only the changes are typed, drafted straight into the mailbox, and the question of how far a tool should be allowed to go before a person confirms
• Projects
• The power of role, backstory and context
• Use case: training recommendations, grounded in the organisation’s own learning catalogue, with a follow-up conversation six months later

The market, then agents

• Which chatbot for which job
• HR-specific tools on the market, each with its weakness said out loud
• What an AI agent is
• The kinds of agent that suit this work
• Use case: an HR policy assistant agent that answers only from the connected policies, quotes the sentence it used, and refers to a person when the policy is silent
• Use case: the same agent given tools, escalating a question by email once the member of staff confirms
• The three levels of building an agent, and where to stop

Data protection, and the honest ending

• What agent building looks like in code, and why you are not the one doing it
• CV screening: the risks, the fixes, and the honest answer
• Cloud versus local hosting
• Running a model with the wifi switched off
• Anonymisation done live and offline
• The closing discussion: screening anonymised CVs against published criteria, what the research actually shows, and where the regulation puts the responsibility

Outcome:

Participants leave with seven use cases they have watched run end to end on real HR work, plus the prompts and set-up to repeat them: a dictated job description with a safety net, an interview question set that improves with every procedure, audience-tailored recruitment posts, personalised candidate letters, development plans grounded in the real catalogue, and a policy assistant agent that knows when to hand over to a person. They also leave with a map of the HR tools on the market, a working method for anonymising data locally, and a defensible position on screening: filtering against published criteria is arguable, ranking people is not, and human oversight is the part that can be held responsible.