Wasting my precious credit

I am developing a web app for my job as a Customer Representative, something relatively simple where I collect data from Discord via bot integration and extract insights like ticket duration, which agent handled the ticket, etc. In the beginning, everything was working very well; conversations with the AI were productive, it solved issues, and identified errors. However, over time, I started noticing that it began spending credits pointlessly, for example, asking me two or three times if it could do something I had already asked it to do. Something like this:

User: Do X, Y, and Z.

Base44: Okay, I will do X, Y, and Z, and to do this, I need to modify such-and-such file.

User: Okay.

Base44: All right, I will proceed and implement X, Y, and Z. Can I continue?

Man, do you understand that when you are charged per interaction and the AI does this, it’s just stealing your credits?

Another thing that has been bothering me A LOT is regarding solutions that don't solve anything. I bring up the problem, explain what is happening, and after a few minutes and credits spent, it says it's resolved, but it’s not. A simple check of the UI is enough to see that it’s exactly the same. Sometimes I have to insist 3 or 4 times before it finally finds the bug and fixes it.

Along these same lines, it has been applying manual fixes; instead of solving the problem, it creates a workaround that only fixes that specific point, but the system is meant to be automatic, which means that in the next update sync, the error will persist for new cases. I’ve already asked it not to do this, to indicate what is happening so I can check. Sometimes the bot lost a permission and the AI doesn't have the data, I don't know, it’s there for that: to show me the error. But even when I explicitly ask it not to work with manual workarounds, in the very next message, it simply applies a manual workaround. It’s completely crazy.

I am happy with the result the app is becoming, but I’m spending credits to correct what the AI should have noticed. It seems absurd to me for it to send "Done, I fixed it," and in the next message, after I say it’s not fixed, it says "You are correct, I am unable to do X, Y, and Z." Why didn't you tell me before that you couldn't do it? It feels like when the problems start getting complex, it just "fakes it" and pretends it’s solving it, but it isn’t.

Taking into account that credits do not accumulate and they are expensive, the use of credits MUST be VERY precise. The AI has to perform every verification before saying that a problem is resolved.