Prompt Chain: Build an AI-Powered Course Development Pipeline

What This Builds

Instead of treating AI as a one-off tool for individual tasks, you'll chain multiple AI prompts together into a repeatable course development pipeline. You start with raw SME content and end with a complete storyboard draft. outline, narration scripts, quiz questions, and scenario descriptions. all in one extended Claude conversation that you run start-to-finish in about 90 minutes instead of 2-3 days.

Prerequisites

• Claude Pro account ($20/month. the 100K context window is essential)
• Google Docs (for organizing and sharing the output)
• At least one piece of SME source content ready to test the pipeline
• Familiarity with your standard storyboard format

The Concept

A prompt chain is like an assembly line for content. Each step builds on the previous one's output:

1. SME content in → Outline out
2. Outline in → Learning objectives out
3. Objectives in → Narration scripts out
4. Narration in → Quiz questions out
5. Quiz + narration in → Scenario activities out

You stay in the same Claude conversation the whole time. Claude remembers what it just wrote, so when you ask for quiz questions "based on the narration you just wrote," it knows exactly what you mean. No copy-pasting back and forth between documents.

Build It Step by Step

Part 1: Set up your pipeline conversation

Open a new Claude conversation → start with a setup message that frames the whole workflow:

I'm going to walk through a full eLearning course development pipeline with you.
We'll go step by step. At each step, I'll give you instructions and you'll produce
output for that step. After I review each step, I'll ask you to proceed to the next.

Course context:
- Topic: [topic]
- Target audience: [audience description]
- Prior knowledge: [what learners already know]
- Reading level: [e.g., 8th grade]
- Number of modules: [e.g., 3]
- Module length target: [e.g., 20 minutes each]
- Assessment type: [e.g., multiple choice, scenario-based]

Confirm you have this context and wait for Step 1.

Part 2: Run Step 1. Content Analysis and Outline

STEP 1: CONTENT ANALYSIS AND OUTLINE

Here is the SME source content:
[paste content]

Analyze this content and produce:
1. A 3-module course outline with:
   - Module title
   - Module learning goal (one sentence: what the learner will be able to DO)
   - 3-5 key topics per module
2. Any significant content gaps where I'll need to go back to the SME

Format as a numbered outline. Keep it scannable — this is for my review, not for learners.

Review the outline → give feedback → Claude adjusts → when satisfied, say "Proceed to Step 2."

Part 3: Run Step 2. Learning Objectives

STEP 2: LEARNING OBJECTIVES

Based on the outline from Step 1, write formal learning objectives for each module.

Requirements:
- 2-3 objectives per module (aim for 2 focused objectives over 3 vague ones)
- Use Bloom's Taxonomy application-level verbs (demonstrate, apply, calculate, distinguish, etc.)
- Format: "By the end of this module, you will be able to [verb] [specific skill]"
- Each objective must be testable — I should be able to write a scenario or quiz question that measures it

Present objectives grouped by module.

Review → refine → proceed to Step 3.

Part 4: Run Step 3. Narration Scripts

STEP 3: NARRATION SCRIPTS

Write narration scripts for all three modules. Requirements:
- Second person ("you", "your")
- Present tense
- [reading level] reading level
- Short paragraphs (3-4 sentences max per paragraph — each paragraph = one slide)
- Each module narration should take approximately [target time] to read aloud at a natural pace
- Start each module with a brief "why this matters" hook
- End each module with a 2-sentence summary of key takeaways

Write Module 1 first, then pause for my review before continuing to Module 2.

Part 5: Run Step 4. Assessments

STEP 4: ASSESSMENTS

Based on the Module 1 narration you just wrote, create the Module 1 assessment. Requirements:
- 8 multiple-choice questions
- Questions distributed across the 3 learning objectives (2-3 per objective)
- Bloom's application level — questions should require using knowledge, not just recalling it
- 4 answer choices per question (A-D)
- Mark the correct answer with ✓
- Write one sentence of specific instructional feedback for each wrong answer

Format as a numbered list with each question, its choices, the correct answer, and feedback clearly labeled.

Repeat for Modules 2 and 3.

Part 6: Run Step 5. Scenario Activity

STEP 5: SCENARIO-BASED ACTIVITY

Create one scenario activity for Module 2 that allows learners to apply what they learned.

Requirements:
- A realistic setting from the learner's actual work environment
- An opening situation (3-4 sentences) that presents a challenge relevant to the module objectives
- Two decision points, each with 3 choices
- Each choice leads to a consequence (positive, negative, or "close but not quite")
- Consequence descriptions: 2-3 sentences explaining what happens and why

Make the wrong choices plausible — not obviously wrong. The scenario should feel like a situation learners actually encounter, not a training exercise.

Part 7: Export to Google Docs

When all steps are complete:

1. Create a new Google Doc named "[Client] - [Course Name] - Storyboard Draft"
2. Copy each step's output from Claude into labeled sections of the document
3. Share with your client or reviewers for markup

Real Example: Safety Training in 90 Minutes

Input: A 22-page OSHA lockout/tagout regulation document + 3-page notes from a plant safety manager interview

Pipeline run:

• Step 1 (10 min): 3-module outline + 2 content gaps identified → sent gap questions to SME
• Step 2 (5 min): 6 learning objectives across 3 modules → approved after minor verb adjustments
• Step 3 (20 min): Full narration for all 3 modules (~1,800 words) → client review suggested 3 small factual edits
• Step 4 (15 min): 24 quiz questions (8 per module) + feedback → all usable, minor wording tweaks
• Step 5 (10 min): 1 realistic scenario with 2 decision points → placed in Module 2

Total AI time: 60 minutes Review and editing: 30 minutes Traditional development time for same output: 2-3 days

What to Do When It Breaks

• Claude loses track of earlier content → Start a new conversation and re-paste the outline as context before running later steps (long conversations can lose early context)
• Narration is too long → Add a word count constraint: "Module 1 narration should be 600 words maximum"
• Quiz questions are too easy (knowledge level only) → Add examples of application-level questions before the prompt: "Here are examples of the difficulty level I want: [paste 2-3 examples]"

Variations

• Simpler version: Run only Steps 1-2 (outline + objectives) and then build narration yourself; use AI for the parts that are most bottlenecked
• Extended version: Add a Step 6 where Claude drafts your kickoff email to the SME with the outline attached and a list of content questions to fill gaps

What to Do Next

• This week: Run the full pipeline on a small course (single module) to test the workflow with real content
• This month: Document your prompt chain as a repeatable process document and time it against your traditional process
• Advanced: Combine this pipeline with the Claude Project (see other Level 4 guide) so every pipeline run automatically applies your client's brand voice and course standards

Advanced guide for Instructional Designer professionals. These techniques use more sophisticated AI features that may require paid subscriptions.