A Custom GPT That Turns Intake Notes into a Job Description, Every Time

What This Builds

Every requisition starts the same way: a hiring manager talks, you type notes, and somewhere between the intake call and the posted job description, the format drifts. One recruiter's JD reads formal, another's reads casual, and a new hire on the team has no idea what "our standard format" even means. This build fixes that by loading your company's actual JD template, tone, and comp-band structure into a Custom GPT once, so every recruiter who uses it gets the same structured output from the same messy notes.

Prerequisites

• Comfortable using ChatGPT to turn raw notes into a first-draft job description (Level 1)
• Plus subscription ($20/month): Custom GPTs require a paid plan
• Your company's JD template, a tone/voice example, and a comp-band reference saved as files you can upload
• Total ongoing cost for this build: $20/month. If your team already pays for ChatGPT Plus for Level 1 drafting, this adds no new subscription.

The Concept

A Custom GPT is a version of ChatGPT that already knows things you'd otherwise have to explain every single time. Instead of pasting your JD template and tone guide into every new chat, you upload them once and write a short set of instructions. From then on, that GPT behaves like a new teammate who read the onboarding docs before their first day: paste in rough notes, and it already knows the format you want back.

Build It Step by Step

Part 1: Gather Your Source Documents

Before touching ChatGPT, collect three files:

• A finished job description your team considers a good example of the standard format
• A short tone note (2-3 sentences: "professional but warm," "direct, no corporate filler," whatever actually describes your employer brand)
• A comp-band reference with ranges by level, redacted of anything you wouldn't want a broader audience seeing if the GPT is ever shared internally

Save these as plain text or PDF files. Keep the comp-band file separate from the JD template so you can choose whether to include it later.

Part 2: Create the Custom GPT

In ChatGPT, open the GPT builder (look for "Explore GPTs" in the sidebar, then "Create"). Give it a clear internal name like "TA Req Intake Assistant" so nobody on the team mistakes it for a general-purpose chatbot. Switch to the Configure tab and paste instructions along these lines:

You are a job description writer for [Company Name]'s Talent Acquisition team.

When a recruiter pastes intake notes from a hiring manager, produce a
structured job description that matches the format and section order in
the uploaded template file. Match the tone described in the tone file:
[paste your 2-3 sentence tone note here].

Always include: role summary, key responsibilities (5-7 bullets), required
qualifications, preferred qualifications, and a closing line about the team.

If the notes are missing a comp range, ask for it rather than guessing.
Never invent responsibilities, requirements, or comp figures that weren't
in the notes or the uploaded reference files.

Flag anything in the notes that reads like it could describe a protected
characteristic (age, gender, family status, and similar) instead of a job
requirement, so the recruiter can review it before posting.

Upload your template and tone files under Knowledge. Leave the comp-band file out unless your team is comfortable with every recruiter who has access to this GPT seeing it, since Knowledge files are visible to anyone who can use the GPT.

Part 3: Test and Refine

Paste in a real (or realistic) set of messy intake notes and check the output against three things: does it match your template's section order, does the tone read right, and did it avoid inventing anything. Adjust the instructions field directly. If it keeps missing the format, name the template file explicitly in the instructions ("always match the structure in JD_Template.docx") rather than assuming it will infer the pattern on its own.

Real Example: Senior Data Analyst Requisition

Setup: TA Req Intake Assistant loaded with the company's standard JD template and a tone note reading "professional but approachable, avoid jargon."

Input: Raw notes from an intake call: "Need someone strong in SQL, has worked with forecasting tools, comfortable presenting to stakeholders, reports to the Analytics Director, hybrid 3 days, band is L4, remote candidates ok if in one of our approved states."

Output: A complete job description with a role summary, seven responsibility bullets built from the notes, a required and preferred qualifications split, and a closing paragraph about the Analytics team, formatted exactly like the uploaded template.

Time saved: The manual version of this step runs 30-60 minutes per requisition. With the GPT doing the first pass, the recruiter's job shrinks to a five-to-ten minute review and edit.

What to Do When It Breaks

• The GPT ignores your template structure → Reference the uploaded file by name directly in the instructions field instead of describing the format in prose. GPTs follow explicit file references more reliably than paraphrased descriptions.
• Knowledge files are long and the GPT seems to skim them → Trim the uploaded template down to the parts that actually define structure. A 20-page style guide gets skimmed; a one-page template gets followed.
• Comp figures or internal notes show up in output shared outside the team → Set the GPT's sharing setting to your organization only, and keep the comp-band file out of Knowledge entirely if anyone outside TA might ever see a shared conversation.
• A new recruiter edits the instructions and breaks it for everyone → Keep one person as the owner of the shared GPT, and save a copy of the working instructions text in a doc outside ChatGPT so you can restore it.

Variations

• Simpler version: Skip the Custom GPT entirely and save the instructions block as a reusable prompt in a doc. Paste it at the top of a new chat each time. Slower, but no paid plan barrier for anyone without Plus.
• Extended version: Build a second Custom GPT loaded with your compliance policy documents for drafting region-specific posting checklists, and link the two so a recruiter can move from JD draft to compliance check in the same session.

What to Do Next

• This week: Run three real requisitions through it and compare the output to what you'd have written manually.
• This month: Share it with the rest of the TA team and collect feedback on where it drifts from the template.
• Advanced: Pair it with a Zapier automation that posts the finished draft straight into your ATS for review instead of copy-pasting.

Advanced guide for Talent Acquisition Specialist professionals. These techniques use more sophisticated AI features that may require paid subscriptions.