top | item 34189292

(no title)

How soon until someone lets you

1) create a series of help docs

2) trains an AI on those help docs

3) solves customer headaches using a chatgpt interface with that training

4) drops you into a human that can help as soon as you ask for it.

My primary issues with "live chat" features are

1) They ask you for stupid stuff (email/name/etc) when you're logged in that are already in the system. This makes things worse somehow.

2) The sheer resistance to handing me over to a human when I've asked for it. If you don't have humans on live chat - that's fine, but at least send it to a help desk.

3) The fact they try to ask me to categorize the issue vs. semantically figuring it out via the request is extra annoying as they will often ask you this AGAIN if they can't initially place you in one of their pre-selected categories.

discuss

SoftTalker|3 years ago

Issue #1 is likely because the "live chat" is a separate product that they bought and linked on the main website, but which is not integrated into the user database or authentication system at all. Therefore the bot needs to ask you for all this identifying information even though you are already logged in.

Meekro|3 years ago

I've worked with products like that on the backend, and they're all perfectly capable of receiving data from your website so that the live chat staffer has all your info automatically show up. Someone was lazy and just didn't do the work to hook it up right.

joelrunyon|3 years ago

I understand that - but they should be able to sync the databases or pass an auth token of some sort.

The fact you have to re-enter it all makes the experience stupid and customers frustrated.

m348e912|3 years ago

This is what I was going to say. Chatbots suck today but with integration of a trained ChatGPT instance they could get substantially better. In some cases it could even better than first-line human support. (I am looking at you AT&T)

smegger001|3 years ago

maybe, but seeing as ChaptGPT already has a habit of making shitup when it doens't know that looks plausible but is wrong I would be worried that it would start sending bad responses make situations worse and not have a obvious way to escalate to a human as its decided it has a 'solved' the problem.