Exactly. Good chat implementations are great; meaningful automated information, fast handover to human operators while also providing them with the relevant information, and all that in the comfort of my browser window is a great experience.
Being stuck in a loop against what's essentially a buggy ELIZA-Clone however, or a system that somehow is incapable of authenticating me even though I've logged into my account on their website, is not.
A lot of the people on HN might be very willing and able to search a FAQ to try and answer their own questions.
But you would not believe how many average people refuse to spend even 10 seconds trying to answer own question and insist on reaching out to a human even when the answer is very obviously readily available. Sometimes this makes sense; there's some scams out there and at least speaking to a native English-speaking person is reassuring when it comes to who you trust your money with. But sometimes some people are just miserable and want to annoy others.
From the other side of the fence - you can base your FAQs on actual questions (to the point of highlighting and bolding the most common) only to keep getting a significant amount of requests with that exact same question multiple times a day.
I can confirm that - at least in my experience - FAQ's are pre-made (and never or rarely updated), when a new service/site/whatever starts, it is logically nonsense that there is even one FAQ.
In theory everything should be clear from the info/documentation/manual, and when it is not so (as it often happens) and something is actually frequently asked, not only an entry in the FAQ list should be added (together with its FGA[0]) but the sheer fact that it is frequently asked should mean that the topic is not clear enough in the info/docs/etc and these should also actually be corrected/updated.
As a "disciplined" user (who actually did read both the FAQ's and the documentation) when you (manage to) contact (via mail/chat or phone) the assistance with a question/doubt there are usually three possibilities:
1) the assistant knows less than you on the topic and cannot answer properly
2) the assistant is competent and manages to answer your question, though with some difficulty (you made an original question)
3) the assistant is competent and answers your question easily because it has already been answered by him/her tens or hundreds of times (your question was not so original but never made it to the published FAQ's)
#3 is the clear sign of a failure in the way FAQ's are managed.
> Acronym for "frequently asked questions"; a list of answers to frequently asked questions that can be presented to a community (be it a forum, Usenet newsgroup, or software user base) so that the same questions need not be asked over and over again. In the entire history of their use, not one has ever been used for its intended purpose.
What the chat should have is something like a chat interface that exposes the FAQ via question and answering, through natural language processing, basically like ChatGPT but it doesn't need to be that advanced.
If the bot can't find the answer, connect to a human. That way, the company's support burden is decreased while still being able to talk to a human if needed.
Being stuck in a loop against what's essentially a buggy ELIZA-Clone however, or a system that somehow is incapable of authenticating me even though I've logged into my account on their website, is not.