Speaker on practical AI adoption in an in-house AI training course for a stainless steel manufacturer in Thailand

How the brief came about

The client wanted a 60-minute training session that would show employees across several job functions how to apply AI to their own work without risking leaks of confidential data. The client held this 1-day course in a meeting room at its own company. The goal was to develop the AI skills of employees at both the head office and the factory.

This session therefore had to teach 2 things at once: how to give AI context, and the rules on what kind of data must not be typed in. If it taught only how to write instructions, employees wanting better answers might enter real company data as context. Participants came from several job functions, so the examples and exercises had to be close to each department's work rather than a single general set of exercises.

Requirements and constraints

Our session had to fit neatly with the other sessions of the course and work for employees across several job functions. The main requirements and constraints were as follows.

• 60 minutes in Session 5 in the afternoon, following the main speaker's session
• According to the schedule, the morning laid the groundwork on AI and had each group make an AI Map (a map of each group's repetitive tasks), while the R-C-F-E framework was taught in the main speaker's preceding session. Our content therefore had to use the same terminology.
• Participants came from several job functions at both the head office and the factory, so the examples and exercises had to cover several departments.
• The content had to align with the rule in the participant materials that confidential company data must not be typed into public AI apps.

Why the worksheet has participants write a Prompt in 2 rounds

The worksheet in the participant materials is designed for writing a Prompt in 2 rounds, so that participants can see how the answer changes when the missing parts are added. We chose this approach because the main problem lay in instructions that were too short, not in the AI app itself. The worksheet refers to the R-C-F-E framework, which according to the schedule was taught in the preceding session, so there was no need to teach the framework again.

The slides for this session gave examples of AI use in factories, such as helping to summarise problems from maintenance reports and helping to review trends in workpiece measurement data. These examples help show what kinds of work AI can be used for. The slides for this session invite participants to try AI chat in a web browser, so they do not need to install any program first. The department-level workshop worksheet in the participant materials sets out a 5-step practice sequence, as follows.

1. 1

   Choose 1 real task

   Each job function group chooses, from its AI Map, a task that is frequently repeated, has clear rules and would not cause serious damage if something went wrong.

2. 2

   Write the first-round Prompt

   Type an instruction into an AI chat app such as ChatGPT or Claude in the way you normally would, then see how usable the answer is.

3. 3

   Find what is missing with R-C-F-E

   Check whether you have told the AI who it should be, what the situation is and what the answer should look like, and whether there is an example to follow.

4. 4

   Write the second-round Prompt

   Add the missing parts and give the instruction again in the same chat, then compare the answers from the 2 rounds to see where they improved.

5. 5

   Prepare a 2-minute presentation

   The worksheet has a group representative cover only 3 things: the chosen task, what was weak about the first Prompt, and what was changed and how that improved the answer.

What materials did Inspektion prepare for this training?

We prepared slides, participant materials, a facilitator guide and backup exercise cards for each group, for use in Session 5. Details of each set are as follows.

• Supporting slides: examples of AI use in factories, and a closing slide encouraging each group to keep well-written Prompts for reuse
• Participant materials: Prompt writing guide, data security rules, Prompt library, workshop worksheet, and post-training rollout plan with a table for recording time saved
• Facilitator guide: the activity sequence for this session and the rules for using the exercise cards
• Backup exercise cards for each group: all fictional data, using names such as "Company A" instead of real names, for groups that cannot think of a task

The participant materials we prepared include a ready-to-use Prompt library for 7 job functions, such as accounting, domestic sales, overseas sales, production planning and purchasing, HR/IT, and production. The 3 data security rules in the participant materials are: do not enter confidential data, every answer must be checked by a person, and AI helps draft but a person does the sending.

Confirmed results from the training day

We were the speaker for Session 5, 60 minutes, in a 1-day in-house AI training course for a stainless steel manufacturer in Thailand in September 2026. The evidence is this session's slides, the supporting materials delivered, and the invoice issued after the training day under the condition of full payment once the training was fully completed. The closing slide encouraged each group to keep well-written Prompts for reuse.

The activity sequence in the worksheet, such as group Prompt writing and the 2-minute presentation per group, is a plan designed in the documents. There is not yet any record confirming that every step was carried out on the day.

There is not yet any post-training measurement that can be cited, because the time saved depends on each department's continued use. The participant materials include a plan for the first 2 weeks and a table for recording the time taken before and after using AI on the same task, so that each department can collect its own results.

What we learned

This work offers 3 principles for training on AI use in organisations.

1. Teach data rules together with how to use AI, not separately later on. A clear rule is to change customer names, prices and costs to fictional data every time before typing, and backup exercises should use only fictional data from the start.
2. A training session that follows another speaker must use the same framework and terminology. In this work, the worksheet was designed to use the AI Map and the R-C-F-E framework, which according to the schedule were in the preceding sessions, so there was no need to teach the framework again.
3. The results of training come after the training day, so measurement tools and a responsible person must be in place from the start. The Prompt library, the plan for the first 2 weeks and the time-recording table help each department collect its own results, but the organisation must appoint someone to compile the results if it wants an overall report.

How can you apply this format to training in your organisation?

You can use the same format if your organisation wants employees across several job functions to start using AI chat in their routine work. The reusable parts are the worksheet sequence, the R-C-F-E framework and the data rules. The parts to adapt are the task examples, the Prompt library and the link with other sessions of the course.

Before holding the training, you should prepare the following.

• Confirm the schedule version actually in use and the role of the speaker in each session, so that the content and terminology connect properly.
• Divide groups by job function and advise the number of people per group in advance, because the exercise cards and presentation time are calculated by the number of groups.
• Make sure participants have accounts and can log in to ChatGPT or Claude before the day, through the app or a web browser, and prepare Wi-Fi, a projector and a microphone.
• Set the organisation's AI use policy, such as what data must not be typed in and whether an enterprise version of the app should be used.
• Collect the tasks each department frequently repeats, to use as exercises in the worksheet.