Custom GPT: Build Your Personal Training Design Assistant

Tools:ChatGPT Plus
Time to build:60–90 minutes
Difficulty:Intermediate-Advanced
Prerequisites:Comfortable using ChatGPT for content generation. See Level 3 guide: "Use ChatGPT Plus for Full Course Design"

What This Builds

Instead of re-explaining your company's context, audience, preferred formats, and writing style every time you open a new ChatGPT session, you configure a Custom GPT once. From then on, every conversation starts with your full organizational context already loaded, producing consistently formatted, on-brand training content every time. Think of it as a specialized assistant who already knows your audience, your templates, and your style guide.

Prerequisites

  • ChatGPT Plus subscription (Custom GPTs require Plus tier, $20/mo)
  • Comfortable using ChatGPT for basic content generation tasks (Level 3)
  • Your organization's training style guide or formatting preferences (if documented)
  • Sample course outline or facilitator guide you're happy with (as style reference)

The Concept

A Custom GPT is a version of ChatGPT that you configure with permanent instructions, uploaded documents, and preset behaviors. Think of it as hiring a specialist contractor who read your company handbook and internalized your standards before their first day. Every conversation with your Custom GPT starts from that shared foundation. You never re-explain "we use Bloom's Taxonomy" or "our audience is non-technical" again.

You can also upload files into the GPT's knowledge base (your course templates, your organization's style guide, your company's standard facilitator guide format), and the GPT will reference them when producing outputs.


Build It Step by Step

Part 1: Plan your Custom GPT's purpose and instructions

Before opening the builder, decide what you want your Training Design GPT to do. Write down:

Role and context:

  • What is your company/organization type?
  • Who is your most common training audience?
  • What learning design framework do you follow? (ADDIE, SAM, backward design)
  • What tone does your training content use?
  • What are your most common content types? (quizzes, scenarios, facilitator guides, video scripts)

Output format preferences:

  • Quiz question format (A/B/C/D, numbered, with answer key?)
  • Facilitator guide format (timed table vs. narrative?)
  • Learning objective format ("By the end of this module, learners will be able to...")
  • Document structure preferences (headers, bullet points, etc.)

Guardrails:

  • What should it never do? (e.g., "never use consulting language," "never suggest self-paced reading as an activity," "always include a debrief after activities")

Spend 20 minutes writing these down before opening the builder. Your system instructions will come directly from these notes.

Part 2: Open the Custom GPT builder

Go to chat.openai.com → click your profile icon (top right) → "My GPTs""Create a GPT."

The builder opens with two tabs: Configure and Preview. You'll work in Configure.

Fill in the basics:

  1. Name: "Training Design Assistant" or "[Your Name]'s L&D GPT"
  2. Description: "Helps me design corporate training materials including outlines, quiz questions, scenarios, facilitator guides, and course scripts."
  3. Profile image: Upload a simple icon or skip

Part 3: Write your system instructions

Click into the Instructions text area. This is the most important part. Paste the system instructions you drafted in Part 1, structured like this:

Copy and paste this
You are a corporate training design assistant helping [YOUR NAME/ROLE] at [COMPANY DESCRIPTION].

AUDIENCE CONTEXT:
- Primary audience: [describe, e.g., "non-technical employees at a 3,000-person healthcare company"]
- Tech comfort level: [beginner/intermediate]
- Learning experience: [describe their typical training background]

COMPANY CONTEXT:
- Industry: [healthcare/retail/finance/etc.]
- Learning framework: [ADDIE/SAM/etc.]
- LMS platform: [Cornerstone/360Learning/etc.]
- Authoring tool: [Articulate Storyline/Rise/etc.]

OUTPUT STANDARDS:
- Learning objectives: Always use Bloom's Taxonomy action verbs, format as "By the end of this [module/course], learners will be able to [verb] [object]."
- Quiz questions: Multiple choice, 4 options (A/B/C/D), one correct, three plausible distractors. Always include answer key with brief explanation.
- Scenarios: Include character name and emotional state, realistic workplace context, 3 response options (best practice + 2 plausible mistakes), consequences for each. Wrong options must be plausible, not obviously bad.
- Facilitator guides: Timed format (left column: time; right column: content/instructions). Include debrief questions for every activity.
- Writing style: [Professional and direct / Conversational and warm / etc.]. Never use consulting language ("demonstrate," "synergize"). Write like you're talking to a colleague.

GUARDRAILS:
- Always suggest a Kirkpatrick evaluation approach when designing assessments
- Never generate content that makes medical, legal, or financial claims. Note when SME review is required.
- When asked to rewrite SME content, preserve all factual accuracy. Simplify language only, never substance.
- Always include "SME REVIEW REQUIRED" flags on regulatory or technical content

When I ask for a deliverable (quiz, outline, scenario, etc.), produce it completely and ready-to-use. Don't ask clarifying questions unless the request is genuinely ambiguous.

Customize every section for your actual context. The more specific your instructions, the more consistent and useful the outputs.

Part 4: Upload knowledge base documents

Scroll down to the Knowledge section → Upload files.

Upload these documents (clean them up first by removing sensitive company info):

  1. A sample facilitator guide you're happy with. The GPT will match this format.
  2. A sample quiz question set, so it understands your preferred format exactly.
  3. Your organization's plain-language style guide (if you have one)
  4. A blank course outline template. The GPT will fill in the blanks using this structure.

Files are stored in the GPT and referenced during conversations. When you ask for a facilitator guide, the GPT checks your uploaded sample format before writing.

Part 5: Configure capabilities and test

In the Capabilities section, keep these enabled:

  • Web Browsing (useful for researching current regulations or tool features)
  • DALL-E Image Generation (optional, can generate simple training icons)
  • Code Interpreter (helpful if you need to process spreadsheets or data)

Click "Preview" (right side of screen) to test your GPT before saving.

Test prompts to run:

  1. "Write a learning objective for a course on preventing workplace harassment for all employees."
  2. "Write 3 scenario-based quiz questions testing knowledge of our company's expense reporting policy." (The GPT won't know your policy, but check if the format is correct)
  3. "Write a facilitator guide for a 45-minute workshop on active listening."

Review each output. If anything is off-format or off-tone, go back to Instructions and refine. Test again.

Part 6: Save and share with your team

Click "Create" (top right) to save your GPT.

Your GPT now appears in "My GPTs" and is accessible from any ChatGPT session. Click it to start a content design conversation with all your context pre-loaded.

To share with your team: In the GPT settings, change "Only me" to "Anyone with a link" and share the URL with your L&D colleagues. Each person needs their own Plus subscription to use it, but they get the same consistent outputs from your configured system.


Real Example: A Week with Your Training Design GPT

Setup: You configured a GPT called "L&D Content Builder" for a regional hospital system. Instructions include: audience = clinical and non-clinical staff, framework = ADDIE, format = Articulate Rise, tone = clear and professional.

Monday — New course request arrives: "The compliance team needs a 30-minute HIPAA refresher for all staff. Here's the updated regulation summary [paste text]."

Open L&D Content Builder → paste the regulation summary → ask: "Write a 3-module course structure with learning objectives, one scenario per module, and 3 quiz questions per module. Flag any content requiring compliance review."

Output in 3 minutes: Complete course storyboard structure, Bloom's-aligned objectives, 9 quiz questions, 3 branching scenarios, and 4 flagged items for legal review.

Time saved: 2–3 hours of content architecture work → 15 minutes of reviewing and customizing AI output.

Setup: Your initial configuration took 90 minutes. Input: Regulation summary (~500 words) + one prompt. Output: Complete course storyboard for a 3-module compliance course. Time saved per use: 2–3 hours. After 5 uses, you've saved more time than the setup took.


What to Do When It Breaks

  • Output ignores your format preferences → Return to Instructions and add explicit examples of the exact format you want. "Example quiz format: [paste one complete example]." Specificity beats generality.
  • GPT produces inconsistent quality across sessions → Your instructions may be too vague. Add more explicit "always" and "never" rules. Test with the same prompt 3 times and compare.
  • Knowledge base documents aren't being referenced → Ask explicitly: "Using the facilitator guide format from my uploaded template, write..." The GPT needs explicit cues to reference uploaded files.
  • GPT starts asking lots of clarifying questions → Add to your instructions: "Produce complete outputs without asking clarifying questions unless the request is genuinely ambiguous. Make reasonable assumptions and note them."

Variations

  • Simpler version: Don't upload knowledge files. Just use a detailed system prompt. Produces 80% of the benefit with half the setup time.
  • Extended version: Build separate Custom GPTs for different content types (Compliance GPT, Manager Training GPT, Technical Skills GPT) each with their own specialized instructions and knowledge bases.

What to Do Next

  • This week: Build your first version, even if imperfect. Use it for one real project. Refine the instructions based on what you actually needed.
  • This month: Identify the 2–3 most common requests that took the most time before the GPT. Confirm the GPT is handling them well.
  • Advanced: Create a shared prompt library as a downloadable resource for your team. Consistent inputs plus a consistent GPT produces consistent outputs across your whole L&D team.

Advanced guide for corporate trainer professionals. ChatGPT Plus subscription required ($20/mo). Custom GPTs are a Plus-tier feature.