Before starting with LINEBOT. Preparations for AI implementation.

Starting next week, we will finally begin building the LINEBOT (= our new AI, "Arai-san").
But before that, there are preparations we need to make as seniors.
Just like when welcoming a new staff member, if we don't set up the environment first, I think it will lead to wasted idle time or cause confusion.

So this time, I have compiled a checklist of preparations actually made before AI implementation.
I organized this from the perspective of "doing this will make implementation smoother," so I think it will be helpful for those who are also considering BOT implementation.

Preparation checklist + what was actually done

1. Checking for gaps in the flow

It's impossible to just dump everything on a new hire and say "please handle it all," right? They will definitely stumble somewhere. Below are examples of "gaps" I anticipated so that Arai-san, the LINEBOT, wouldn't end up in that situation.

• Multiple questions mixed in a long sentence... might result in unanswered questions

• Only an attachment is sent... might not be able to read it

• No information in the FAQ to refer to... might not be able to provide an accurate answer

So, I wrote down rules for how to behave in exceptional cases.

Even if it's not perfect, having minimum standards for "what to do when in trouble" provides a sense of security.

2. Initial FAQ setup

It's unreasonable to throw a new hire into the field without giving them a textbook.
I picked out frequently asked questions and made them the initial FAQ.

• "When will the delivery be?"

• "Where can I check the invoice?"

• "How do I perform a re-delivery?"

Even just this ensures that "Arai-san can answer immediately from the first day."

3. Slack notification design

Arai-san will signal, "I answered up to this point, but this is impossible."

I set up a two-tiered notification system.

I focused on the balance of "being able to see everything, but only getting notified about things that require action."

4. Approval flow

When Arai-san suggests, "Can I add this answer to the FAQ?", who gives the go-ahead? If this isn't decided, the answers won't get updated.
Regarding that, we decided to set a weekly time to review FAQ candidates and designate a responsible team. The idea is that a mention will be sent to the team on Slack saying, "Please approve this."

5. Draft of operational rules

Questions not in the FAQ must be forwarded to Slack -> The person in charge replies -> Use actions like pressing a dedicated stamp as a trigger to add it to FAQ candidates if necessary.
Reviewing FAQ candidates will be a weekly check.

Just deciding to "review it once a week for now" significantly reduces hesitation in operations.

Summary

AI implementation is not something that "automatically becomes convenient once you start."
Just like when welcoming a new staff member, setting up the work environment first is the most important thing.

Prepare as a senior so you don't panic after Arai-san arrives.
I feel that this extra effort prevents confusion after implementation and becomes the foundation for smoothly running improvements.

After this, so that Arai-san can start working, I plan to finish setting up a simple flow in n8n and verifying connectivity (which is like "preparing a desk" in a real office) before finally having her start working in earnest.