How to Design an AI Training Program — A Practical Guide to Department-Specific Curricula and Impact Measurement

This is Hamamoto from TIMEWELL.

"We want to raise the AI literacy floor across the whole company, but we don't know how to design the training." This is a question I'm being asked more and more from HR and learning and development teams.

The context is clear. Nomura Research Institute data shows that 57.7% of Japanese companies have already deployed generative AI, but 70.3% of those companies cite "insufficient employee literacy and skills" as a key challenge. The tools are in place; the people who can use them aren't. Leave that gap unaddressed and the AI investment becomes pure sunk cost.

---

Why "Ad-Hoc Training" Fails

Let me lay out the most common failure patterns first. If any of these sound familiar, it's time to rethink your design.

Failure Pattern	Example	Root Cause
Lecture-heavy	Three hours of AI theory, no hands-on practice	Gap between knowledge and application
One-size-fits-all	Same curriculum for sales and engineering	Ignores differences in departmental work
Event-only	Annual training with no follow-up	No plan to reinforce learning
Goal-free	"Deepen AI understanding" as the objective	Unmeasurable goal
Fully outsourced	Training company handles everything, no internal know-how built	No internalization strategy

The common problem across all of these: "running the training" has become the goal. The actual goal is "employees can use AI in their daily work" — not the training event itself. Confuse the two and you end up with an annual ritual that consumes budget and collects satisfaction surveys.

---

Five Steps to Curriculum Design

Step 1: Understand the Current Skill Level

Before designing anything, get an accurate read on where employees actually are. Use a company-wide survey or sampling to identify the following:

Level	Definition	Corresponding State
Level 0	No experience	Has never used an AI tool
Level 1	Has tried it	Has experimented with ChatGPT or similar personally
Level 2	Using at work	Uses AI regularly in daily business tasks
Level 3	Applied use	Can design prompts and integrate AI into workflows
Level 4	Promoter	Can coach others and plan AI utilization initiatives

According to IPA's "DX Trends 2025," 85.1% of Japanese companies feel a shortage of DX talent. In my experience, at most companies, Level 0 to 1 accounts for 60–70% of the workforce. Building a curriculum without this baseline picture means guessing in the dark — and landing either too advanced or too basic.

Step 2: Set Department-Specific Learning Targets

Rather than company-wide uniform goals, define what each department should be able to do after training. This is the core of curriculum design.

Department	Target	What They'll Be Able to Do
Sales	Level 2 → 3	Use AI to draft proposals; apply AI to customer analysis
Engineering	Level 2 → 4	Integrate AI into code review; design test automation
Admin (HR, Finance, General Affairs)	Level 1 → 2	Use AI daily for form drafting, data compilation assistance
Marketing	Level 2 → 3	Generate content ideas; streamline market research with AI
Corporate Planning	Level 1 → 3	Assist with management data analysis; use AI for meeting material preparation

It's obvious that sales and engineering need very different skills — but what gets overlooked is the admin function. HR, finance, and general affairs are often assumed to be "AI-irrelevant," but they handle large volumes of tasks where AI delivers immediate value: template documents, FAQ responses, and similar routine work.

Step 3: Build the Curriculum

Work backward from the learning targets to construct the curriculum. Here are practical configurations by department.

---

Sales Curriculum (4 sessions × 2 hours each)

Session	Theme	Content	Hands-On
1	AI basics and sales applications	How generative AI works; where it fits in the sales process	Try competitive research using ChatGPT
2	Efficient proposal creation	Prompt design fundamentals; managing output quality	Draft a proposal for an actual deal
3	Customer analysis and insight discovery	Analyzing meeting history; generating customer need hypotheses	Analysis exercise using CRM data
4	Practical workshop	Integrating AI into your team's actual workflow	Each participant presents their business improvement plan

Engineering Curriculum (4 sessions × 2 hours each)

Session	Theme	Content	Hands-On
1	AI-driven development fundamentals	Coding assistance AI; overview of test automation	Try code generation with GitHub Copilot or similar
2	Prompt engineering	Technical prompt design; context control	Implement complex requirements using AI
3	Code review and quality management	AI-powered code review; security checks	Experience automating review of existing code
4	Integration into the development workflow	Embedding in CI/CD; setting team usage guidelines	Design how to integrate AI into your team's development flow

Admin Curriculum (3 sessions × 2 hours each)

Session	Theme	Content	Hands-On
1	AI basics and admin applications	Basic generative AI operation; information security precautions	Draft emails and meeting minutes
2	Routine task efficiency	Document creation, data organization, FAQ response automation	Create actual department forms using AI
3	Practical application and reinforcement	How to integrate into workflow; understanding internal guidelines	Create a one-month AI utilization plan

A note on the most important element of any curriculum: the hands-on material. Use the participants' actual data and actual work — not hypothetical case studies. For sales, use real deal data. For admin, use the reports they actually produce every month. Without the experience of "this actually helps my work," behavior won't change after training.

Step 4: Set Up the Delivery Structure

Once the curriculum is finalized, decide how to deliver it.

Format	Advantages	Disadvantages	Best For
In-person group session	Active Q&A, builds team cohesion	Hard to schedule, higher cost	Kickoffs, workshops
Online (synchronous)	Location-independent, can be recorded for review	Harder to sustain attention	Lecture portions, foundational content
E-learning (asynchronous)	Self-paced	Harder to maintain motivation	Knowledge input
On-the-job training	Directly tied to real work, high retention	High burden on the trainer	Advanced skill development

My recommended combination: e-learning first for foundational knowledge, then a group session for hands-on practice, followed by OJT for reinforcement. Spending group session time on lecture content is a waste. Move the theory to pre-work e-learning and use the group time for hands-on exercises.

Step 5: Establish Operating Rules and Follow-Up Structure

Training doesn't end when the session does. The follow-up structure that sustains skill development until it's embedded in daily work is what actually determines whether training succeeds.

Examples of follow-up activities:

• Weekly 15-minute retrospectives where each person shares an example of how they used AI that week
• A dedicated Slack or team chat channel for questions and information sharing
• Monthly skill checks with a quick test to verify comprehension
• An internal AI utilization contest at the three-month mark, collecting and recognizing business improvement examples

Honestly, the follow-up structure has more impact on outcomes than the training content itself. Three hours of training won't change anyone. Three months of continuous practice will.

---

Measuring Outcomes

A framework for measuring training effectiveness and reporting results to executive leadership.

Four-Level Evaluation Using the Kirkpatrick Model

Level	What's Evaluated	How to Measure	When to Measure
Level 1: Reaction	Participant satisfaction	Post-training survey	Immediately after training
Level 2: Learning	Knowledge and skill acquisition	Comprehension test, skills assessment	At training completion
Level 3: Behavior	Application to work	Usage rate survey, manager assessment	1–3 months after
Level 4: Results	Contribution to business outcomes	KPI comparison (time reduction rate, error rate, etc.)	3–6 months after

Most companies stop at Level 1 satisfaction surveys. But what executives want to know is Level 3 to 4 — "what changed in how people actually work." High satisfaction with no change in practice means the training investment isn't being recovered.

Specific KPI Examples

KPI	How to Calculate	Benchmark Target
AI utilization rate	% of employees using AI at least once a week	70%+ at three months post-training
Work time reduction	Before/after comparison of time spent on target tasks	20–30% reduction in target work
Output quality	Error rate, customer satisfaction, number of review comments	10–20% improvement
Internal case count	Number of AI utilization/improvement examples reported	At least 1 per department per month
Skill level advancement	Pre/post skill assessment score comparison	Average improvement of 1+ level

CloudAce research shows that 80.2% of companies that set KPIs for their AI utilization initiatives achieved their targets. The reverse is also true: companies without KPIs struggle to produce measurable results. Build the measurement framework in from the start — it's a prerequisite.

---

Leveraging Subsidies

AI training may qualify for Japan's "Employee Career Development Assistance Subsidy" (人材開発支援助成金). This Ministry of Health, Labour and Welfare program subsidizes a portion of the expenses and wages associated with employee vocational training.

Key requirements include:

• Submitting a training plan in advance and obtaining approval from the regional labor bureau
• The training must be conducted as off-JT (outside of regular duties)
• Meeting the minimum training hours requirement (10 hours or more, etc.)
• Filing an application for the subsidy after training completion

The eligibility requirements and subsidy rates change annually, so check the Ministry of Health, Labour and Welfare website or your nearest Hello Work office for the latest information. The application process is admittedly cumbersome, but the program can significantly reduce training costs — it's worth using.

---

Curriculum Design Checklist

A checklist of items that tend to be missed at the design stage.

During the design phase:

• Have you conducted a company-wide skill level assessment?
• Have you defined learning targets by department?
• Is the curriculum balanced between knowledge and practice?
• Does hands-on practice use actual company data and real work?
• Have you incorporated information security guidelines?

During the delivery phase:

• Have you distributed pre-learning materials?
• Have you selected and prepared the trainer (internal or external)?
• Have you secured participants' schedules?
• Have you prepared the hands-on environment (tool accounts, etc.)?

During the follow-up phase:

• Have you planned post-training follow-up activities?
• Have you decided on outcome measurement KPIs and measurement timing?
• Have you set a reporting schedule for executive leadership?
• Have you designed an improvement cycle for the next round of training?

---

Summary

Training should be evaluated not on "whether it happened" but on "whether it produced results." Building the measurement framework into the design from the start is the only way to justify training investment.

As your next action: start with a company-wide skill level assessment. Once you know where people are, the question of which departments need what level of curriculum answers itself. Planning a "company-wide AI training" without that assessment means designing a program that won't actually connect with the front line.

---

TIMEWELL's WARP provides end-to-end support for corporate AI training program design, delivery, and outcome measurement — with curricula directly tied to practical application. WARP NEXT is a long-term accompaniment model where the program is updated monthly to reflect the latest AI developments. WARP BASIC is a customized short-format program with department-specific curriculum options. Reach out even if you're at the stage of thinking through the overall training design.

Learn more about WARP