This article is a revised and expanded version of one I published on a different account around May 2026, now republished on this account.
In my previous article, “I Started Real Estate Crowdfunding,” I summarized the situation from when I started real estate crowdfunding up to the time of publication.
From here, I would like to delve a little deeper into the “use of AI in considering real estate crowdfunding” that I touched upon in the previous article.
Honestly, without the help of AI, I think my investment decisions would still be a mess. There are many things I don’t understand even when looking at project introductions, and the truth is I don’t have the time or luxury to read through every single detail.
However, as long as I can provide the necessary information to the AI, it will read through all the documents and break down the following points for me.
-
The pros and cons of the project
-
The difficulty of the project (risks and points of caution)
-
How well it aligns with the investment policy I provided in advance
There is a proverb that says “use whatever you can, even your parents,” but even if it’s not your parents, if there is an AI you can use, you should make full use of it.
However, you are the one who makes the final decision.
For example, even if you provide the exact same information, the answers may differ depending on the AI. Also, it is not uncommon for an AI to misunderstand and return an incorrect answer. What’s tricky is that it’s not a mistake that anyone would notice at a glance, but rather it sneaks errors into the correct information. When this happens, it’s not straightforward.
That is precisely why I keep the AI’s answers strictly as reference material, and I make sure I am in a position where I can confidently say “I made the final decision myself” based on various information, including the AI’s advice. This is my basic stance.
◇
The AI tools I use on a daily basis are the following three types.
-
Copilot (it’s just there if you use Windows)
-
Gemini (it’s just there if you use Google services)
-
ChatGPT (the well-known catalyst for the spread of AI)
I didn’t consciously choose these three; they just happened to be close at hand and easy to use. And the reason is that all of them can be used for free to some extent. They are all well-known names.
At first, I consulted Copilot about projects because it was the only one I could use for work and I was most accustomed to it. However, when I suddenly decided to consult ChatGPT and Gemini about the same project, I noticed that their answers were different.
Even if the conclusion itself was the same, the detailed opinions differed, they analyzed from completely different angles, and sometimes their views were split right down the middle between “it’s okay to invest” and “you shouldn’t invest.”
This was a very interesting discovery. Just as the view changes for humans depending on their perspective, it is natural for AI to produce different results because their structures and mechanisms differ.
I have realized that utilizing multiple AIs to hear diverse opinions is extremely effective for investment decision-making. For this reason, I continue to operate by asking the same questions to three different AIs.
◇
As you continue to converse with an AI, knowledge (insights and context) accumulates there. It is, so to speak, your own personal know-how.
However, there is a limit to the capacity that can be stored in a single chat. As the conversation grows longer, the AI begins to forget the older content. Also, depending on the type of AI, some will continue the conversation but gradually become slower in processing, while others will suddenly cut off the conversation once a certain amount of data is reached.
In any case, it becomes necessary to periodically reset (create a new) chat.
The problem here is that when you start a new chat, past memories are not carried over. Honestly, it is a hassle to explain my premises and policies from scratch every time and have the AI relearn them.
So, I asked the AI itself how to deal with this.
In conclusion, the solution is to create “handover data” in advance and have the AI read it at the beginning of a new chat.
Even if you cannot carry over everything perfectly, you can sufficiently carry over important points and prerequisites. You can even consult the AI on how to create this “handover document” by asking, “I want to create a handover document like this; how should I structure it?” and it will create a base draft based on the conversation history so far. After that, you just need to continue the dialogue with the AI to make revisions and improve the accuracy.
I personally create and use handover data that covers the following items.
-
Purpose of the handover document
-
Role expected of the AI
-
Current investment status
-
Portfolio policy
-
Project evaluation policy
-
Other operational rules
If changes occur in these contents, I update them as needed and have the AI read them again every time I start a new chat.
By the way, in my case, I add instructions for specific AIs in addition to this. Specifically, for some reason, Copilot’s responses were in a bulleted list format, which made it very difficult to read, so I added instructions to prevent that. In reality, even after adding those instructions, it continued to use bullet points, so while we were brainstorming, I got angry and wrote something to the effect of, “I wrote in the instructions to stop using bullet points, can you not read? How many times do I have to tell you that it is very hard for a human to read?” and it finally understood and switched to a normal writing style. Well, this is just an aside (laughs).
◇
Now, given the explanation so far, you might be wondering, “What kind of handover document are you specifically creating?” However, after careful consideration, I have decided to refrain from disclosing the details of the handover document here.
I believe that the content of a handover document should differ from person to person depending on their investment policy and risk tolerance. I have no objection to showing my own settings, but they are by no means perfect. Rather than using them as a direct reference, it is undoubtedly more effective for each person to create their own in a way that works best for them.
This may have sounded a bit arrogant, but I hope to have another opportunity to talk in detail about the specific creation process and operational methods of this handover document.
