Write a Job Description That Isn't Boilerplate
The 15-minute version, plus the one review step most people skip
The task and where Claude helps
Most job descriptions are copies of the last version, which was copied from the one before it. You end up with a wall of generic requirements, generic applicants, and good candidates screening themselves out.
Think about the last “HR Business Partner” posting you saw that could have been for any company in any industry. Or the “Data Analyst” posting that listed every tool under the sun, then hired someone for three core reports and headcount dashboards. That is the pattern you are trying to break.
Claude is useful here because a JD is a structured writing task with a clear shape. It can turn raw bullets into clearer responsibilities, strip filler phrases, and help you keep must-haves separate from nice-to-haves. It is not a final authority on what the role should be and it does not decide who you hire. That judgment stays with you and the hiring manager. Use Claude to draft and pressure-test; keep the decisions with humans.
The copy-paste prompt
You are helping me write a job description. I will give you the basics.
Write a clear, specific JD that avoids generic filler and corporate cliche.
Role: [job title]
Team and reporting line: [team / manager]
Why this role exists right now: [the real business reason]
The 4 to 6 real tasks this person will own: [list]
Must-have skills or experience: [list]
Nice-to-have: [list]
Location / work model: [onsite / hybrid / remote]
Rules:
- No cliches like "rockstar," "wear many hats," or "fast-paced environment."
- Separate true must-haves from nice-to-haves; do not inflate requirements.
- Use plain, direct language a real person would say out loud.
- Keep responsibilities concrete and outcome-focused.
- Flag any requirement that might unnecessarily narrow the candidate pool.One realistic example
Here is what that prompt looks like when you actually use it.
Role: Senior People Analytics Partner
Team and reporting line: People Analytics, reports to Director of Talent Strategy
Why this role exists right now: We are moving from reactive reporting to forward-looking workforce planning and need someone to own our headcount and attrition models instead of pulling ad-hoc reports.
The 4 to 6 real tasks this person will own:
Own monthly people metrics dashboard for two business units.
Build and maintain workforce planning models.
Partner with HRBPs on scenario analysis for hiring and restructuring.
Run quarterly deep-dives on attrition.
Maintain people data documentation and basic data quality checks.
Must-have skills or experience:
Experience working with people data and dashboards.
Strong Excel or BI skills.
Comfort partnering with HR and business leaders.
Ability to explain analysis to non-technical leaders.
Nice-to-have:
Experience with your specific BI tool.
Prior work in a SaaS environment.
Experience with compensation data.
Location / work model: Hybrid, three days a week in our Pompano Beach office.
From that input, Claude can produce a JD that sounds like a job a real person could do and want, not an analytics “rockstar” in a generic “fast-paced environment.”
What to feed it
Better inputs produce better drafts. Start with the basics above, then add:
Your existing JD for the role, even if you know it is stale. Claude can strip the filler and keep the core.
A sentence or two on team culture, for example, “We are a small HR team that values direct communication, quick iteration, and practical solutions over slide decks.”
The actual comp band for the role (the range, not your internal philosophy) if you are comfortable, so the tone matches the level.
Any JD whose tone you like and the instruction “match this tone.”
The “why this role exists right now” line does more work than any other input. It is what turns a list of tasks into a job worth wanting.
Examples:
“We are opening two new regional hubs in the next twelve months and need a recruiter who can own those markets end to end, from sourcing to offer close.”
“Our HRIS implementation is complete, but nobody owns the data quality and reporting layer. This role exists to clean up the basics and then build reporting we can trust.”
Each of those gives the role a reason to exist beyond “support business objectives.”
What good output looks like
A usable draft has three markers.
First, the responsibilities read as outcomes. You want “Own the monthly close for US payroll and partner with Finance on reconciliations” instead of “Responsible for various payroll duties.” The candidate should be able to picture what they will own on a normal Tuesday.
Second, the must-have list is short and defensible. For a mid-level payroll role, “Two or more years of payroll experience, comfort with US multi-state payroll, experience working with Finance or Accounting teams” is defendable. “Seven or more years of experience, advanced degree in Accounting, prior experience in our exact industry, expert-level skills in three different systems” is not, and it will quietly narrow your pool.
Third, it sounds like a person wrote it. “You will spend most of your time building and explaining reports for leaders who are trying to make better headcount decisions” reads human. “You will use data-driven insights to inform strategic decision-making in a fast-paced environment” reads like the last hundred postings you have seen.
If the draft still reads like every other JD, your input was too thin. Add the “why now” reason and one line of team context, then regenerate.
The trap to avoid
Three risks matter here.
The first is requirement inflation. AI has seen long lists of “must-haves” and will gladly give you another one. Imagine a People Ops Generalist role that truly needs basic HR process experience and comfort talking to managers. The draft comes back with three named HRIS platforms, a certification requirement, and seven years of experience because that is the pattern in the data. Every unnecessary requirement narrows your pool and can introduce adverse impact. Cut the list to what the job needs on day one.
The second is biased or exclusionary language. Watch for coded terms like “digital native,” “recent graduate,” and “strong cultural fit,” and for unnecessary degree requirements in roles that can be done without a four-year degree. A frontline HR coordinator job that requires a bachelor’s “in Business or related field” may screen out experienced candidates with community college, military, or non-traditional backgrounds who have done the work. Ask Claude directly, “Flag any language that could deter qualified candidates or create adverse impact,” then use your own judgment on what it surfaces.
The third is oversharing in the prompt. Do not paste anything you would not want outside your walls. That means no “we are quietly exiting our current Head of HR,” no internal comp philosophy that would surprise employees if it leaked, and no unannounced reorg context. A JD is a public document. Treat the draft process the same way.
One obvious guardrail sits under all of this. A draft is a draft. The hiring manager and, where relevant, legal or compliance still review before it posts.
The level-up move
Once the JD is solid, chain the workflow.
From the final JD, ask Claude: “Generate a structured interview guide: five competency-based questions mapped to the must-have skills, with what a strong answer looks like for each.”
For the Senior People Analytics Partner example, you might see questions like:
“Tell me about a time you built a workforce planning model that changed a headcount decision. What did you build and what changed?”
“Walk me through a dashboard you owned. Who used it, what decisions did it support, and how did you know it was working?”
Now your JD and your interview process draw from the same source of truth. You are assessing candidates on the same skills you wrote into the role instead of letting each interviewer improvise.
This is No. 2 of The HR Practitioner’s Claude Playbook. New workflows weekly. Questions or requests: info@futurefluenthr.ai.