Hiring Without an HR Department
How to Write a Job Description with AI in 15 Minutes
Writing a job description with AI comes down to three moves: paste your messy notes about the role, let the AI produce a structured first draft, then edit in the honesty about your company that no tool can supply. This guide walks through the whole process, with four copy-paste prompts.
1-on-1 with Pete Enestrom, on your actual work
Written by Pete Enestrom
Yale & Columbia, ex-Microsoft & Intel — 1-on-1 AI coaching for executives
Your coach’s background
At a company of fifty or five hundred people with no HR department, a job description gets written by whoever has an hour free: the founder, an office manager, the person the new hire will report to. The result is usually a rushed document copied from an old posting. The posting matters more than it looks—a job description is the first filter on who applies, and honest postings attract better-matched candidates.
AI changes the starting point, not the standard. A tool like Claude or ChatGPT takes your roughest notes about a role and returns a clean, conventionally structured draft in under a minute. What the tool cannot supply is the truth about your company—what the work actually is, what the last person in the seat found hard, what your culture rewards. That part still comes from you, and it is the part candidates are screening for.
This guide is the process I teach hiring managers at mid-market companies: how to get a strong AI draft of a job description, how to tighten the draft so the posting screens for the right things, and how to keep the format consistent as the team grows. Four copy-paste prompts do the heavy lifting. Everything here works in the free versions of Claude and ChatGPT (Anthropic's pricing page lists Claude's free tier).
What can AI actually do for a job description?
A job description AI draft is strong on structure. Give Claude or ChatGPT fragments—half-sentences, a task list, a complaint about the last hire—and the tool returns the conventional shape candidates expect: a role summary, ordered responsibilities, requirements, and a closing note about the company. The draft arrives in seconds, reads cleanly, and frees the hiring manager to spend the saved hour on judgment instead of formatting.
The limits of an AI job description draft show up in the content, not the formatting. An AI tool does not know that this operations role exists because the last person burned out covering two jobs, or that your warehouse team will quietly ignore a manager who has never worked a shift. Generic input produces a generic posting, and a generic posting attracts a generic pool. The quality ceiling on an AI job description is set by the honesty of the notes you feed it.
How do you write a job description with AI, step by step?
Plan on about fifteen minutes of your own time once the notes exist. The prompts mentioned in each step appear in full in the next section.
Dump the messy reality of the role into a note
Open a blank note and write for ten minutes: what this person will actually do week to week, who they work with, what broke the last time this seat was empty, and what a good first year looks like. Do not tidy it. Fragments and half-sentences are fine—the mess is the raw material, and the AI's first job is organizing it.
Ask for a structured first draft
Paste the notes into Claude or ChatGPT with the role-from-notes prompt below. The tool returns a conventional draft: summary, responsibilities, requirements, company note. Read the draft as a candidate would, and mark every line that does not match the real job.
Tighten the requirements until each one earns its place
Run the requirements-tightening prompt below. The goal is a short “must have on day one” list and a separate “can learn here” list. Every requirement you cut widens the pool; every vague requirement you keep screens out people who would have been strong hires.
Add the honesty the AI can't supply
This pass is yours, not the tool's. Name the real manager and team. Include one honest trade-off—the role is new, the process is not built yet, the travel is real. Candidates who opt out after reading an honest posting were never going to stay; the ones who opt in arrive with their eyes open.
Interview against the posting, then save everything
Use the interview-questions prompt below to build the interview loop directly from the final text—if a line in the job description cannot be tested in an interview, the line does not belong in the posting. Then save the posting and the prompts in a shared folder, so the next hire starts from the same format instead of a blank page.
Four prompts that do the heavy lifting
Copy these verbatim, replace the [brackets] with your material, and adjust the closing lines to taste. The prompt is the easy part—the honesty of your notes is what makes the output worth posting.
I'm the [your title] at [one line about your company]. I'm hiring a [role title]. Here are my raw notes about what this person will actually do, who they work with, and what a good first year looks like: [paste your notes—fragments and half-sentences are fine]. Write a job description with: a two-sentence summary of the role in plain language, 6–8 responsibilities ordered by how the person will actually spend their time, and a short paragraph on what success looks like after the first year. No buzzwords—no “fast-paced environment,” no “rockstar,” no “world-class.”
The messier the notes, the better the draft. Tidy input usually means you have already started writing a generic posting in your head.
Here is the requirements section of a job description I'm drafting: [paste]. Do three things: (1) split the list into “must have on day one” and “can learn in the first six months,” (2) flag any requirement that isn't strictly necessary for the work I described, and (3) rewrite each remaining requirement as one plain sentence. Ask me up to three questions if any requirement is ambiguous.
Long requirement lists quietly shrink the applicant pool—many strong candidates only apply when they meet every line. Cutting is recruiting strategy, not lowering standards.
Here is the final job description for a [role title] at my company: [paste]. Write ten interview questions that directly test this posting—six about the responsibilities listed and four about how the person works with a team. For each question, add one sentence describing what a strong answer sounds like, so a hiring manager without an HR department can score answers consistently across candidates.
This is also a quality check on the posting itself: any line you cannot interview against is decoration, and decoration misleads candidates.
I'm writing a job description for a [role title], roughly [seniority level], based in [location or “remote”]. Using only general conventions, draft: a one-line statement of how senior this role is, the experience that seniority usually implies, and placeholder text for the pay range reading “[range to be confirmed]”. Do not invent a salary figure—I will add the real range myself. Then flag any phrasing in the draft that could raise legal or compliance questions in a job posting, so I know what to check before publishing.
Keep your actual compensation bands out of the tool. Real pay ranges come from your comp plan, and pay-transparency rules in some states and cities require them in postings—have whoever owns compliance confirm the final wording.
What does AI get wrong about job descriptions?
The first failure mode of an AI-drafted posting is inflation. Left alone, an AI tool pads a posting with boilerplate—“dynamic team,” “competitive salary,” “must be a self-starter”—because that language dominates the postings the tool learned from. Boilerplate inflates the word count and deflates the signal: a candidate learns nothing about your company from a paragraph that could sit under any logo. Cut every sentence a competitor could publish unchanged.
The second failure mode of an AI draft is confident invention. An AI job description draft will happily specify years of experience, software names, and certifications you never mentioned. Some of those guesses will be plausible and wrong for your seat. Read the requirements line by line and delete anything you did not put there on purpose—the tool is guessing at a typical role, and you are hiring for a specific one.
The five-minute check before the posting goes live
Six questions. Run them on every AI-drafted job description, and the posting will be more honest than most of what it competes against.
- Would we genuinely reject an otherwise strong candidate who lacked each requirement on this list?
- Does the posting name the real team and manager, and include at least one honest trade-off about the role?
- Did every salary figure come from our actual compensation plan rather than from the AI's suggestions?
- Has whoever owns compliance confirmed the pay-transparency and equal-opportunity wording for the locations where this role can be filled?
- Did we keep confidential material—comp bands, org plans, customer names—out of the AI tool entirely?
- Does the format match our other live postings, so candidates see one company and not three voices?
How do you keep job descriptions consistent as the team grows?
Consistency is a solved problem once the first honest posting exists. Save one skeleton—the same sections in the same order—and one prompt file alongside it. Each new role starts the same way: the hiring manager dumps messy notes, the AI produces the structured draft in the house format, and one named person does the honesty pass before anything goes live. Candidates comparing two of your postings should see one company speaking, not a different voice per department.
The hiring manager always owns the content; the AI only owns the formatting. That division holds as volume grows—a company opening ten roles a quarter has no excuse for ten different posting styles, and no excuse for letting the tool write the parts only a person can know.
What changes when someone teaches you this?
Everything on this page runs this week, alone, for free—and for a hiring manager filling one seat, self-teaching is enough. The pattern changes when hiring is continuous: the question stops being how to draft one posting and becomes how your whole team writes, screens, and interviews with the same discipline.
That shift—from drafting one posting to running a whole hiring process—is the point where a human teacher—not an AI coach bot, not another template pack—earns the fee. In 1-on-1 coaching we open your real hiring workflow, build the prompts and review habits around your company, and get your managers fluent enough to run the process without me. This guide is the map; coaching is walking it on your terrain.
1-on-1 coaching
Want a teacher, not another tutorial?
Guides get you started. A 1-on-1 session gets you fluent—built on your actual work, at your pace.
Your coach
Coached by Pete Enestrom
Yale and Columbia grad, former Microsoft and Intel, and a venture-backed exited founder. Pete has spent the last four and a half years going deep on every major AI tool, and he teaches the way operators learn: on your real work, at your pace, with nothing assumed.
Common questions
Straight answers, the way I'd give them across a table.
Keep reading
Guide
AI for Performance Reviews: A Manager's Workflow That Works
A manager's workflow for using AI on performance reviews: turn a year of raw notes into fair, specific first drafts, then apply the judgment only a manager has.
ReadGuide
ChatGPT for Work: A Practical Guide for Non-Technical Executives
How non-technical executives use ChatGPT at work: drafting, summarizing, brainstorming, notes into action lists, and meeting prep—with six copy-paste prompts.
ReadFor Your Role
AI for Executives: What Real Fluency Looks Like for Business Leaders
AI for executives means fluency, not a certificate: what fluency looks like for a CEO or president, and the five leadership conversations AI now touches.
ReadBeyond coaching
When the workflow needs to be encoded, not just learned — that's what our team at Zaigo builds.