Honestly, this is an excellent use for AI. So long as responses are limited to the context of the customer support document database. Which is why it’s astounding that no one is talking about it.
Force customers to use it for 5 minutes and you cut your call volume by 70% (RTFM).
Ideally they could take those savings and invest it in their human support providers but of course that’s very unlikely in most cases.
That would be awesome. But it sounds like the llm has some access to customer accounts and call tools. Prompt injection remains a fundamentally unsolved problem, so I’m curious about how Intel plans to handle that.
ya ok… this is how any of my calls go when recordings from ANY company answer and it won’t change, but it’ll involve me driving to the company to physically demand assistance for whatever crap they caused. almost ALL of my issues, involve a human brain to think and can’t be fixed by some air trash or scripted messages.
‘hello and thank you for calling some garbage corporation that doesn’t care about you in the least. please listen fully as our options have changed 12 years ago but we’re too lazy and cheap to change this recording’
0#0#0#0#0#0#0#0#0#0#0#0#0#0#0
‘that option is not valid…for customer serv…’
human please… human… customer service. human
‘please hold while I connect your call to a live representative who only ever says exactly what is in the script, and can’t think for themselves, please subscribe to our garbage services in the meantime and remember… you are disposable if we can profit’
please hold while I connect your call to a live representative who only ever says exactly what is in the script
I mean, what you have here is a complaint with customer service in general. I’d bet the AI I’ve described is far more helpful than some person half way across the world that doesn’t even speak English.
I’ve only ever seen and used it 1 time and it was awesome.
Honestly, this is an excellent use for AI. So long as responses are limited to the context of the customer support document database. Which is why it’s astounding that no one is talking about it.
Force customers to use it for 5 minutes and you cut your call volume by 70% (RTFM).
Ideally they could take those savings and invest it in their human support providers but of course that’s very unlikely in most cases.
That would be awesome. But it sounds like the llm has some access to customer accounts and call tools. Prompt injection remains a fundamentally unsolved problem, so I’m curious about how Intel plans to handle that.
They don’t plan to handle it lol
Read access or write access?
ya ok… this is how any of my calls go when recordings from ANY company answer and it won’t change, but it’ll involve me driving to the company to physically demand assistance for whatever crap they caused. almost ALL of my issues, involve a human brain to think and can’t be fixed by some air trash or scripted messages.
‘hello and thank you for calling some garbage corporation that doesn’t care about you in the least. please listen fully as our options have changed 12 years ago but we’re too lazy and cheap to change this recording’
0#0#0#0#0#0#0#0#0#0#0#0#0#0#0
‘that option is not valid…for customer serv…’
human please… human… customer service. human
‘please hold while I connect your call to a live representative who only ever says exactly what is in the script, and can’t think for themselves, please subscribe to our garbage services in the meantime and remember… you are disposable if we can profit’
I mean, what you have here is a complaint with customer service in general. I’d bet the AI I’ve described is far more helpful than some person half way across the world that doesn’t even speak English.
I’ve only ever seen and used it 1 time and it was awesome.
But usually the elephant in the room for GenAI is ethics.
I hope it isnt here.