Sector primer: AI in HR

If you work in HR — recruiting, people ops, L&D, internal comms — there’s a good chance your week looks different than it did six months ago, even if you haven’t named the change. Here’s a structured tour of what’s actually shifted, what hasn’t, and what to be careful about.

What’s already changed

1. Resume screening. This is where most HR teams started. Tools that can read 200 resumes in 90 seconds and surface the 15 most relevant for human review have become standard. The catch: how the tool defines “relevant” is opaque, and it’s where bias quietly enters. Ask your vendor to show you their evaluation methodology, and rotate human review on the rejected pile, not just the shortlist.

2. Interview prep. Generating role-specific behavioral questions, summarizing candidate background, drafting offer letters in tone-consistent ways. All faster. All also requires a careful pass — AI happily invents a candidate’s job history if you let it.

3. Policy drafting. First-pass employee handbook updates, leave-policy language, return-to-office memos. Drafting time has roughly halved. The legal review still matters — and the AI-drafted version often sounds polished while missing local nuances. (Quebec employers, hi.)

4. Internal Q&A. Some HR teams now run an internal AI assistant that can answer “how many vacation days do I have left, what’s the parental leave policy, do I need to fill out a form for X.” It cuts your inbound by 30–60% if it’s set up well. It causes new headaches if it’s set up poorly.

What hasn’t changed

The hard part of HR is still the same. Conversations about performance, conflict, layoffs, accommodations, harassment, and trust. AI doesn’t help with any of that — and pretending it does is one of the ways HR teams lose their judgment muscle.

The other thing that hasn’t changed: the values work. What kind of company you want to be, who gets included, how decisions get made. AI can summarize a survey. It cannot tell you what to do about it.

What to be careful about

Three places we’d push hard:

• Don’t run AI through candidate communications without disclosure. Candidates increasingly know. A canned “thank you for your interest” is fine. A personalized rejection generated by AI that’s pretending to be human is a trust break.
• Don’t optimize for the metric you can measure. AI tools optimize for what you tell them to. If “time-to-fill” is the only metric, you’ll fill faster — and miss long-term fit. Keep the human metrics in the loop.
• Don’t outsource your policy voice. AI-drafted policies often read identically across companies. Your culture has a voice. Use AI for the first draft. Then write over it.

What we’re listening for

If you’re in HR, we want to hear: where has AI been actually useful, and where has it caused problems you didn’t expect? Email us — confidential, of course. We’re collecting these for an upcoming episode.