Does designing an automatic interaction based on user intent a good thing?

We have this flow in our interface where user can unlink their accounts that are connected to the application. We have a button to unlink their bank, but we also have an option where user can manually unlink each account. The process here is quite tricky because they need to click the “submit” button once they are okay with their changes, reason being they need to enter 2FA,and we do not want them to enter 2FA for each change or unlink that they do for each account if needed. Sorry for the bad sketch but this is how it is currently laid out:

In cases that user tries to manually unlink each account and does not use the “unlink bank” option, the suggested interaction is when user clicks the last account with the “Unlink”, the system will detect their intent to unlink all accounts and the bank and will show a confirmation dialog if they want to unlink the bank:

It is a good interaction as it detects the intent of the user and it removes one step which is clicking the “update” button, however my concern is that clicking on the last “unlink” button and then showing a confirmation dialog after might be something unexpected? I’m not sure if the modal interrupts the user work flow since it is detecting that it wants to unlink all accounts which equals to unlinking the bank and one less click of clicking the update button, while the modal offers a second chance whether to cancel or not. Is there an article that would suggest this is a bad idea?

We cannot do user testing at the moment that is why this is challenging. But I’d like to get your thoughts or if you have any articles you could share that relates to this topic, that would be really helpful.