Most HR work is writing the same documents over and over: job posts, offer letters, policy updates, the careful email that turns down a candidate without burning a bridge. That's exactly the kind of drafting a good chat model does in seconds, and where it saves an HR team the most time.
The trick isn't asking for "a job description." It's giving the model your specifics — the level, the team, the must-have skills, your tone — so the draft comes back usable instead of generic. The prompts below are written that way. Paste one, swap in your details, edit the result.
Every prompt below works in a free Chatbot App account. No credit card to start.
Try these prompts freeGeneric JD templates over-list requirements and scare off good applicants. Feed the model the seniority, the actual day-to-day, and your three non-negotiables, and ask it to keep the list short.
Write a job description for a mid-level Backend Engineer on a 6-person platform team. Day-to-day: maintaining a Node/Postgres API, on-call one week in six, mentoring one junior. Must-haves: 3+ years production backend, comfort with SQL, async written communication. Keep requirements under 7 bullets, drop corporate filler, and write the intro paragraph in a plain, direct voice.
Paste an existing policy and ask for a plain-English rewrite at a set reading level. Useful for handbooks employees actually read.
Rewrite this remote-work policy in plain English at an 8th-grade reading level. Keep every rule, but cut legalese and use short sentences. Flag anything ambiguous that a manager would need to clarify. [paste policy]
Give it the role rubric and a batch of resumes, and have it score against your criteria with reasons. Treat it as a first pass a human reviews, never the final decision.
Here is a scoring rubric for a Customer Success Manager role (weight: relevant CS experience 40%, SaaS exposure 25%, written clarity 20%, ownership signals 15%). Score each resume 1-5 per criterion with a one-line reason, then give a total. Do not infer gender, age, or background. [paste rubric + resumes]
Candidate rejections, PIP framing, a heads-up about a benefits change. Ask for two tone variants so you can pick the one that fits the relationship.
Write a rejection email to a final-round candidate we liked but passed on for a more senior hire. Warm, specific, leaves the door open for future roles. Give me one version that is 4 sentences and one that is 2 sentences.
Never paste anything that identifies a real employee or candidate into a tool you haven't vetted for data handling — salary, health, performance details. Anonymize first, or keep those drafts in a self-hosted setup. For screening, the model is a first filter; a person makes the call.
No, and treating it that way gets you in trouble. It replaces the blank page — drafting, summarizing, rewriting. Judgment calls, employee relations, and anything legally sensitive stay with people. Think of it as a fast junior writer who never gets tired of the tenth job post.
Only if you know how the provider handles it. As a rule, strip names and identifiers before pasting, and check whether your plan trains on your inputs. For sensitive records, anonymize the data first.
They look generic only if you prompt generically. The fix is feeding real specifics — team size, actual tasks, your three real requirements — so the output is grounded in your details instead of boilerplate.
Unlimited AI access for $19.99/month, or start free with 100 messages a day.