Skip to main content

AI for HR: Writing Job Descriptions That Aren't Boring - Digital Compliance Academy

Most job descriptions are terrible. They are long, generic, and biased. Here is how HR teams can use Claude to write descriptions that actually attract talent.

Jon McGreevy September 20, 2025 3 min read
HR Recruitment Bias Claude

I have a theory: nobody has actually read a job description from start to finish since 2015.

Candidates scan the salary, the title, and maybe the first bullet point. Why? Because most JDs are walls of corporate text, copied and pasted from a template written ten years ago, full of phrases like “passionate self-starter” and “dynamic environment.”

For HR teams, writing these documents is a chore. It’s a “Grudge Task.” And when we do Grudge Tasks, we do them badly.

This is the perfect use case for AI. But not to just generic text generation. We can use tools like Claude 3.5 Sonnet to audit, rewrite, and improve our recruiting assets.

The “Bias audit” Prompt

We know that unconscious bias creeps into job descriptions.

  • Masculine-coded words (“Aggressive,” “Dominant,” “Rockstar”) can deter female applicants.
  • Ageist language (“Digital Native,” “Recent Grad”) can deter experienced hires.

Claude is an excellent neutral auditor.

The Prompt:

“Here is a draft job description for a Senior Developer. Please audit it for unconscious bias. Specifically, identify any gender-coded language or ableist terms. Then, suggest 3 alternative phrases for each flag that would make the role more inclusive.”

The “Day in the Life” Rewrite

Most JDs tell you what the company wants (Requirements), not what the human gets (Experience). Use AI to flip the perspective.

The Prompt:

“I am going to paste a list of responsibilities for a Customer Support Manager. based only on these tasks, please write a ‘Day in the Life’ summary. Write it in the second person (‘You will start your day by…’). Make it sound challenging but rewarding. Avoid corporate jargon.”

Using AI to Screen (The Danger Zone)

A warning: Do not use public AI to screen CVs.

I repeat: Do not upload applicant CVs into ChatGPT or Claude.

  1. Privacy: That is PII (Personally Identifiable Information). You are breaching GDPR. (See our GDPR Guide).
  2. Bias: AI models can inherit racial or gender biases from their training data. If you ask it to “Rank these candidates,” it might rank them based on names or schools rather than skills.

The Recruitment “Agent” Workflow

Instead of replacing the human decision-maker, use AI to prepare the interview materials.

Step 1: The Competency Map

“We are interviewing for a Head of Sales. The key competency is ‘Resilience’. Please give me 5 behavioural interview questions that test for resilience, along with a ‘Good Answer vs Bad Answer’ scorecard.”

Step 2: The Role Play

“I want to practice the interview. You act as the candidate. I will ask you questions. Give me answers that are technically correct but lack specific examples, so I can practice probing for more detail.”

Why HR Needs to Lead This

HR is often the “Department of No” when it comes to new tech. But AI offers a chance to be the “Department of Flow.”

By using AI to strip away the admin of writing policies and JDs, HR professionals can get back to the part of the job that actually matters: talking to humans.

Tools Recommendation:

  • Writing: Use Claude. It is less “salesy” than ChatGPT and writes more empathetic, human-sounding copy.
  • Policy Analysis: Use Gemini. It can handle your entire 50-page Employee Handbook and answer questions like “Does our current policy cover AI usage?”