What AI actually looks like for an HR team
HR involves a lot of writing, reviewing, and communicating. Here's where Claude saves real time — and where to be careful.
HR teams are writers and communicators. Job descriptions, offer letters, policies, onboarding materials, performance review templates, difficult conversation frameworks — the function produces a huge volume of text. Claude is genuinely useful here, with some important caveats.
What HR teams use Claude for
Job descriptions. This is often the first thing HR teams try, and it works. Give Claude the role title, team context, key responsibilities, and must-have experience. Claude produces a structured draft. The main edit required: make sure the language sounds like your company, not like every other job description on the internet. A style guide uploaded to a Claude Project helps with this.
One specific use: inclusion review. Ask Claude to review a draft job description for language that might unintentionally discourage applications from underrepresented groups. It's not a substitute for professional DEI guidance, but it catches obvious things.
Interview question banks. Give Claude the role requirements and ask for 20 structured interview questions organised by competency. Then edit to keep the ones that reflect how your team actually evaluates candidates. Building a question bank from scratch takes hours; editing a draft takes 20 minutes.
Policy drafts. New policies — remote work, AI usage, expenses, parental leave top-ups — need drafting. Claude produces a solid first draft from a brief description of what you want the policy to cover. You then edit for your legal jurisdiction, your culture, and what your leadership has actually approved. Never ship a policy draft without review — this is one of the cases where the human review step is not optional.
Onboarding documentation. Week-one checklists, team guides, tool access runbooks, culture docs. Claude can draft all of these. The editing work is catching the company-specific details it can't know: the specific Slack channels, the way your standup works, who to ask about what.
Communication drafts for difficult situations. Redundancy communications, performance improvement plan language, internal announcements about sensitive changes. Claude can draft these in a neutral, professional tone. The requirement: a human with judgment reviews every word before anything goes to an employee. Claude can help you not start from a blank page; it cannot replace HR judgment in sensitive situations.
Summarising feedback. Paste 20 performance review responses and ask Claude to identify common themes — what people appreciate, what concerns come up repeatedly. Useful for identifying team-level patterns without having to manually code qualitative data.
Where to be careful
Anything that affects employment decisions. Claude helps with drafts and summaries; a human makes every decision. There's meaningful legal and ethical risk in using AI outputs as the basis for hiring, firing, or performance decisions without appropriate human oversight.
Jurisdiction-specific legal requirements. Claude knows a lot about employment law in general but doesn't know your specific jurisdiction's requirements, your company's negotiated terms, or your legal counsel's guidance. All policy and contract drafts need legal review.
Employee data. Don't paste individual employee records, performance data, or personal information into Claude.ai. If you're building an internal HR tool with the API, your data handling practices need to comply with your privacy policies.
The setup that makes it practical
A Project with:
- Your company values and culture description
- Your tone guidelines for HR communications (formal vs. conversational?)
- Your current headcount and growth stage (helps Claude calibrate role descriptions)
- Any standard templates you want Claude to follow
The goal: HR team members can get a useful first draft for almost any communication task without explaining company context each time.