I Used AI to Build a 4-Hour Training Course in 1 Hour: 15 Deliverables, $15 Total Cost

GoodTech AI · 6 min read

 

Table of Contents

1. Why Training Course Creation Takes So Long
2. 8 AI Agents Built My Course Materials
3. 15 Deliverables: What Actually Came Out
4. Outsourcing vs. AI: The Numbers Side by Side
5. What AI Cannot Do (Yet)
6. FAQ

 

Last week, I needed to build a 4-hour AI training course from scratch.

 

A proposal, curriculum, 56-slide deck, speaker scripts, handouts, and hands-on exercises. Under normal circumstances, this is a two-week project. Outsourcing it to an external instructor in Korea typically costs $1,500 to $3,500.

 

I have spent 18 years in HR at a Korean manufacturing company, designing training programs for over 1,000 employees. The content itself was never the hard part -- it was the sheer volume of materials that consumed all my time.

 

This time, I tried something different. I handed the entire job to an AI multi-agent system called Claude Code. The result: 15 deliverables in 1 hour, at a total cost of $15.

Why Training Course Creation Takes So Long

In Korea, 7 out of 10 companies have already implemented AI training programs (Hunet, 2025). Demand for AI and digital transformation education surged 290% year-over-year (Korean Standards Association, 2025).

 

Training demand is exploding, but course creation is still manual.

 

Here is the reality I have lived for nearly two decades: when you commission an external instructor, expect 2-3 preliminary meetings, 3-4 rounds of draft feedback, and 1-2 weeks until final sign-off. Cost: $1,500 to $3,500 including instructor fees and design work. Rush jobs push timelines even further.

 

The irony every training designer knows: the time spent creating the training often exceeds the training itself.

---

8 AI Agents Built My Course Materials

The method I used was Claude Code's multi-agent system. Instead of one AI doing everything, 8 specialized agents worked in sequence, each handling a different role.

 

Step	Agent	Role
1	Researcher	Searched the web 12 times for current data and case studies
2	Curriculum Designer	Module-by-module time allocation with theory, demo, and practice segments
3	Critic (Round 1)	Reviewed curriculum for logical gaps and missing components
4	Script Writer	Speaker notes for all 56 slides
5	Critic (Round 2)	Re-examined script flow and time distribution
6	Slide Builder	Generated PPTX file with 12 auto-created images
7	Handout Creator	Produced learner materials and practice prompt collection
8	Quality Assessor	Scored output across 10 criteria (100-point scale)

The key design choice: the Critic agent appears twice. Round 1 catches structural issues in the curriculum. Round 2 catches quality issues in the scripts. It mirrors how a human reviewer would work -- AI checking AI.

 

I was skeptical at first. "What difference would AI-on-AI review make?" But the Critic's feedback was genuinely sharp: it flagged modules with insufficient practice time and transitions between slides that felt abrupt.

---

15 Deliverables: What Actually Came Out

One hour later, here is what appeared on my computer:

 

#	Deliverable	Description	Scale
1	Training Proposal	Overview, objectives, curriculum, expected outcomes	7,300 chars
2	Curriculum Design Doc	Module-by-module timing, activities, transitions	29,000 chars
3	Speaker Script	Full narration for all 56 slides	62,000 chars
4	Slide Guide	Per-slide layout and design specifications	49,000 chars
5	PPTX File	Presentation-ready PowerPoint	5.3 MB
6	Learner Handout	Practice guide + reference materials	13,000 chars
7-11	Practice Materials (5)	Prompt templates, sample spreadsheets, exercises	6 files
12-15	Slide Images	AI-generated illustrations	12 images

The 56-slide speaker script alone was 62,000 characters -- roughly 155 pages of manuscript. Writing that by hand would take days.

---

Outsourcing vs. AI: The Numbers Side by Side

Based on my experience managing training budgets in Korean corporate environments:

 

Category	Manual / Outsourced	AI-Created
Time	1-2 weeks	~1 hour
Cost	$1,500-$3,500	$15 (Claude API usage)
Deliverables	PPT + 2-3 handouts	15 total (proposal to exercises)
Revisions	Re-negotiation + extra fees	Re-run with modified prompt, near-zero cost
Quality Check	1-2 internal reviews	2 AI critic rounds + 10-criteria assessment (90/100)

Research from Wharton and Menlo Ventures (2025) found that every $1 invested in generative AI returns an average of $3.70 in value. In Korea, one major conglomerate reported a 31% reduction in annual overtime hours after implementing AI in their training workflow (Comento HRD Forum, 2025).

 

---

What AI Cannot Do (Yet)

Let me be direct: you cannot use AI-generated training materials as-is.

 

Here is what I found in this specific project. Some case studies did not match the target industry context. Practice time allocations were unrealistic in a few modules. And certain sentences had that unmistakable "AI wrote this" quality.

 

"The core challenge is not the technology itself, but how to apply it to actual work and scale it across organizations." -- Korean Standards Association, HRD Trends 2026

 

AI creates the first draft. The strategic direction, instructional design expertise, and understanding of your actual learners -- that remains your job. Even after two rounds of AI critic review and a 90/100 quality score, I still made manual edits before the final version.

 

---

Frequently Asked Questions

Q. Do I need coding skills to use this multi-agent approach?

A. Claude Code runs in a terminal, but you interact with it using natural language. Basic comfort with a command line is helpful, but no programming knowledge is required.

Q. Can this method work for topics other than AI training?

A. Yes. The multi-agent structure is topic-agnostic. It works for compliance training, onboarding programs, leadership development, technical skills -- any structured course format.

Q. How does the quality compare to a human-created course?

A. The AI output scored 90/100 on its own quality assessment, but it required human editing for industry-specific context, tone, and realistic time allocation. Think of it as a strong first draft that an experienced designer refines.

---

When the Draft Is Free, You Can Focus on What Matters

Building a 4-hour training course in 1 hour means something specific: more time for the work that actually matters -- designing the learning experience, understanding your audience, and making the content genuinely useful.

 

Before discovering this method, I spent most of my time producing materials and had little left for the core question: "How should this actually be taught?"

 

If you design training for a living, try running your next course through AI. It will not be perfect. But having a complete first draft changes everything -- because refining is always faster than creating from scratch.

 

One step at a time. I hope this article is the first one.

 

GoodTech AI

Technology for the benefit of people