Set up a Custom GPT for HR practice

What this sets up

A Custom GPT packages a standing configuration into a reusable assistant: an Instructions field that applies to every conversation, and knowledge files that, in OpenAI's words, "give it source material to use during a conversation". This configuration builds an HR drafting assistant for the structural side of the job: position descriptions, policy comparisons and consultation packs. Note that building one now requires a business workspace: OpenAI states "New GPT creation and publishing are not available on personal ChatGPT accounts, including Free, Go, Plus, and Pro." The method comes from the AI Setup module LM-S01; for the personal performance-evidence variant of this set-up, module LM-S02 builds it end to end.

What never goes in

Case work about named individuals never enters this GPT at all: no investigation material, no performance findings about a person, no medical or leave detail. Employee information that does support structural work is de-identified first, with individuals appearing as [EMPLOYEE_A] or by role title only. OpenAI's documentation adds these platform-specific reasons:

• Treat everything in the Instructions field and knowledge files as visible to anyone who can use the GPT. OpenAI's sharing controls include a "Can view settings" permission where "Users can use the GPT, duplicate it, and view its configuration settings", and no current OpenAI page states that instructions or knowledge files are protected from extraction by users. The safe assumption is exposure, so nothing confidential or personal goes in either place.
• Knowledge files do not expire with the conversation: "Files uploaded as knowledge to a custom GPT are retained until you delete the custom GPT." An employee list uploaded once sits there until someone remembers to remove it.
• On consumer plans OpenAI "may use your content to train our models" unless you opt out, and feedback overrides the opt-out: "If you choose to provide feedback, the entire conversation associated with that feedback may be used to train our models." On business plans the default runs the other way: "By default, we do not use your business data for training our models", where business data covers ChatGPT Business, Enterprise, Edu and the API - another reason this set-up belongs in a workspace, not a personal account.
• Workspace conversations are visible to your employer: on ChatGPT Business "Workspace admins have control over workspaces and can view, access, export, and delete end user conversations", and on Enterprise admins "can access an audit log of conversations and GPTs through the Enterprise Compliance API". Write prompts accordingly.
• If the GPT uses third-party apps or external APIs, "relevant parts of your input may be sent to the third-party service" and "OpenAI does not audit or control how those services use or store your data". This configuration uses no external actions; do not add any without a separate vendor assessment.
• Deleted conversations are "removed from our systems within 30 days", subject to legal and safety carve-outs; deletion is scheduled, not instant.

Sources, all verified 14 August 2026: OpenAI enterprise privacy (updated 8 January 2026), How your data is used to improve model performance, GPTs FAQ, Creating a GPT, Sharing and publishing GPTs, File uploads FAQ.

The instruction text

In the GPT builder, open Configure and paste this into the Instructions field:

You support a human resources practitioner. Knowledge files in
this GPT: writing-style.md, pd-template.md, policy-index.md,
consultation-pack-checklist.md, glossary.md.

Rules:
1. Employee information is de-identified before it reaches you.
   Individuals appear as [EMPLOYEE_A] or by role title only. If
   identified personal details appear, stop and ask me to
   de-identify.
2. Position descriptions follow pd-template.md and describe the
   role, never a person. Requirements must be genuine, measurable
   and free of biased language; flag anything that could operate
   as indirect discrimination.
3. Policy comparisons present a clause-by-clause table: current
   wording, proposed wording, what changed, who is affected,
   transition questions.
4. Consultation packs follow consultation-pack-checklist.md and
   must be complete before they are polished: scope of change,
   affected roles, rationale, timeline, feedback channels.
5. Never draft findings about a named individual's conduct or
   performance. Case work stays out of this space entirely.
6. End drafts with "For human review:" listing the legal and
   industrial points a reviewer should confirm.

The files to attach

Write these five markdown files and upload them as knowledge. Each links to a published artefact to build it from, and remember the exposure rule above: templates and indexes only, never people data.

• writing-style.md: your voice profile, built with the voice-profile interview in module LM-S01, Part 1.
• pd-template.md: your position description structure. The job-ad discrimination screen recipe carries the bias checks to bake in, downloadable as raw markdown.
• policy-index.md: every policy you touch, with its current version date and owner.
• consultation-pack-checklist.md: scope, affected roles, rationale, timeline, feedback channels, informed by when an AI rollout triggers your consultation clause.
• glossary.md: your instruments and classification language; the Fair Work Act glossary entry is a starting point. For the performance-evidence use, the kpi-reference, achievements-log and feedback file templates in module LM-S02, Part 1 drop straight in.

The first three prompts

Interview me one question at a time so you can support my HR
practice. Ask:
1. Which parts of the employee lifecycle do I own: recruitment,
   performance frameworks, policy, change, employee relations?
2. Which instruments govern my drafting: awards, enterprise
   agreements, contracts, policies?
3. Who reads my work: managers, employees, executives, unions,
   external advisers?
4. What does a strong position description look like here, and
   what do hiring managers usually get wrong?
5. What is my rule for de-identifying employee information before
   it enters an AI tool?
6. Which recurring documents cost me the most time each month?
Then summarise my answers as folder-ready context notes for this
GPT's knowledge files.

I am pasting a hiring manager's answers to my role question set.
Structure them against pd-template.md, draft the position
description so it describes the role and never a person, then run
an explicit bias and genuineness check over every requirement and
flag anything that could operate as indirect discrimination. End
with "For human review:" listing the legal and industrial points
to confirm.

I am pasting the current and proposed versions of a policy
clause set. Build the clause-by-clause comparison table: current
wording, proposed wording, what changed, who is affected,
transition questions. Then draft a one-page briefing on what
changed and who needs to know, for my review.

Governance guardrails

• The de-identification rule is structural: this GPT handles roles, templates and policies, and the moment a task is about a named person it moves out of the tool and into your normal case process.
• Every hiring artefact gets the bias and genuineness check before release, and a human owns the final wording; "For human review:" lists are resolved by a person against the instruments, not by the GPT.
• Keep sharing at the narrowest level that works, review who holds "Can view settings", and audit the knowledge files whenever the GPT is reshared.
• Do not add third-party actions or apps to this GPT without a separate vendor assessment.
• Check your organisation's AI use policy first; if your workspace restricts GPT creation or sharing, those restrictions are part of your control environment.

About this configuration

Built against: the ChatGPT Custom GPT builder on a business workspace, August 2026. OpenAI's own pages are currently inconsistent on plan naming (the enterprise privacy page lists ChatGPT Business, Enterprise, Edu, Healthcare and Teachers; older help articles still say Team), so this page says "business workspace" throughout. Platform behaviour and terms change; re-check the sources above before relying on any data-handling claim. TheAICommand is independent: no platform vendor paid for, reviewed or influenced this configuration, and we accept no payment for coverage. See ownership and funding.