Iro AIWhat you'll be able to do
- Draft an inclusive job description with biased or exclusionary wording flagged for you
- Turn a policy outline into a clear first draft written for the people who'll read it
- Build a fair, role-relevant interview question bank in minutes
- Summarize feedback and survey comments without amplifying one loud voice
- Find the right tone for a sensitive message before you have to send it
Inside the path
A focused set of five-minute lessons — each one ends with a hands-on exercise, not a quiz you can guess.
Inclusive job descriptions 6 min
Draft JDs that fit the role and the audience, and have AI flag wording that could deter qualified candidates.
Policies people can actually read 5 min
Turn a rough outline into a clear, plain-English policy draft written for the employees it affects.
Fair interview question banks 5 min
Generate role-relevant, consistent questions and screen out ones that invite bias.
Summarize feedback fairly 6 min
Condense survey comments and reviews into balanced themes instead of the loudest quote.
Tone for sensitive comms 5 min
Get the wording right for hard conversations — while keeping names and private details out of the tool.
Try a sample exercise
This is the kind of card you'd practice inside Iro — you do the thinking, then get feedback.
◆ Sample exercise · Prompt practice
You need to post a senior backend engineer role at a 30-person fintech and want AI to draft the job description — without introducing wording that quietly narrows your candidate pool.
Your task: Pick the prompt that gets a strong, fair draft you can stand behind.
- "Write a job description for a software engineer."
- "Draft a job description for a senior backend engineer at a 30-person fintech startup. Tone: warm but professional. Must-haves: 5+ years, strong Python, payments experience. Keep it inclusive — flag any wording that could deter women, older candidates, or people from non-traditional backgrounds, and suggest neutral alternatives. Leave salary out for now."
- "Here's our candidate spreadsheet with names, ages, and current salaries — write JDs and rank who to interview."
- "Write the most demanding job description you can so weak applicants don't bother applying."
See why the second prompt wins
The winning prompt sets a clear audience and context (senior backend, 30-person fintech), specifies tone (warm but professional), pins down the real must-haves so the draft is accurate, and — most importantly — builds in a bias check: it asks the model to flag wording that could deter specific groups and to suggest neutral alternatives. The generic "software engineer" prompt gives you filler you'll rewrite anyway. The spreadsheet option makes two serious mistakes at once: it leaks confidential employee PII into a general tool and hands candidate screening to AI, which a human must own for fairness and compliance. And "most demanding possible" bakes in exclusionary bias by design. In Iro you'd write your own version and get feedback on audience, tone, and where bias and confidentiality need a human check.
AI drafts the words; a person owns the decisions
Most HR work is language: job descriptions, policies, offer letters, onboarding guides, feedback summaries, and the careful messages that go with hard moments. That's exactly what AI is good at — taking a blank page and giving you a solid first draft in your audience's language. Used this way it clears hours of writing so you can spend your judgment where it matters.
What AI must never do is make the call about a person. It shouldn't rank candidates, decide who's underperforming, or determine an outcome, because those decisions carry fairness, legal, and human weight that a text-prediction tool can't hold. The line is clean: AI drafts and structures; a person reviews, applies policy, and stays accountable for anything that affects someone's job.
The two risks AI won't manage for you
Two problems are specific to HR, and AI won't catch them unless you make it — or unless you simply don't create them:
- Bias: models learn from human text, so they can quietly reproduce gendered or age-coded language and unfair patterns. Turn that around by asking AI to flag biased or exclusionary wording and suggest neutral alternatives — and never let it screen or rank people.
- Confidentiality: employee names, PII, health details, compensation, and investigation notes should not go into a general AI tool. Draft with placeholders and roles instead of real identities, and follow your organization's data policy.
Handle those two and AI becomes a genuine time-saver across hiring, policy, and onboarding — without putting a person's privacy or a fair process at risk. This is a productivity tool; the responsibility for fair, compliant people decisions stays with you.