I'm curious to see how empowered the LLM customer service of the future is. Are they going to be allowed to process refunds? People will find out how to convince the LLM you do meet the requirements for a refund.
And if you check the requirements using something deterministic (last order was within last 2 weeks, customer has a high loyalty points score, they're subscribed to some premium service) then why do you need an agent at all? The real point of the agent when it comes to something like that is reducing churn by making the process harder. The LLM isn't working for you, just like the human agent isn't working for you.
I do think this sort of thing could be good for digging up information though, as technical support before you get to a technician, could be good. Probably less aggravating than "Before we connect you, please turn on and off the rout-... you pressed 1, that is an invalid option. Before we connect you, please turn..."
Most requests are very routine and are basic exchanges of information. Think "how do I reset my password?".
Don't see why those can't be automated.
Eventually I don't see why an AI shouldn't process refunds. Its an infinitely measurable process and you always know the risk involved (the amount). So it's trivial to decide if you think the AI should handle it or send it off to a human for approval. I would not be surprised if there were already companies experimenting with stuff like this. Widespread adoption will probably take time.
And if you check the requirements using something deterministic (last order was within last 2 weeks, customer has a high loyalty points score, they're subscribed to some premium service) then why do you need an agent at all? The real point of the agent when it comes to something like that is reducing churn by making the process harder. The LLM isn't working for you, just like the human agent isn't working for you.
I do think this sort of thing could be good for digging up information though, as technical support before you get to a technician, could be good. Probably less aggravating than "Before we connect you, please turn on and off the rout-... you pressed 1, that is an invalid option. Before we connect you, please turn..."