We've all been there. We've all gone to a website that probably rhymes with "grammars on" in need of a human customer services representative and had to battle through the AI assistant first.
"Speak to a human"
"Wait, no, I'm sure I can help you :):)"
"No, speak to a human"
"Come on, give me a chance!"
"Okay I need to return this product outside of the standard 30 day window"
"To do this, you'll need to speak to a human. Please wait while I transfer you".
Sigh
Has anybody ever actually used one of these AI assistants and been satisfied with the outcome? Seriously, I'd like to get the executives who signed off on these things in a room and force them to use their own tools. It used to be that companies were praised for their excellent customer services, but now you rarely hear about it.
Don't get me wrong, I don't hate AI or automation; I just strongly believe that it has its place, and that front line customer services isn't it. Ever since machines came about, people who performed manual tasks have been scared of losing their jobs. "Don't worry" they said. "Machines won't replace you, they'll assist you." they said. Or my personal favourite "Machines and AI will make it so that humans won't have to work - we'll be able to relax, make art, and enjoy life". Little did we know that the AI would be making the art while humans get to do the mundane stuff that AI can't do (yet).
What should be happening in the workplace is that the day to day mundane stuff should be taken away from people who really shouldn't have to be doing data entry all day, so they can focus on the important things like actually answering questions from customers. What's actually happening is staff numbers are being reduced because the AI is taking the front line and redirecting the "outliers" to a smaller team of humans. The end result is that realistically, nobody is happy. Staff members aren't happy because any day now, their jobs could be taken by a mediocre AI assistant. Customers aren't happy because it's a pain to get in touch with anyone who can actually help. Can the big bosses really be happy when staff and customer satisfaction is through the floor?
Let's think about how LLMs actually work. Rather than "learning" things, and actually understanding what they're asked and then giving an informed response, they're trained on an insanely large dataset of mostly real human interactions that show it how questions are formed, and answers are given. So if there are 3,000 instances on the web of someone asking "What colour is a London bus?" and 3,000 answers saying "Red", chances are that the LLM will spit out that answe to the same question. However, the training dataset doesn't take sarcasm or good-old-fashioned jokes into account, which is why AI assistants that might rhyme with "lemon eye" suggest things like adding glue to pizza sauce, or that sharks are older than the moon.
Going further into the training, we can even think about the "Dead Internet Theory" which basically suggests that a hell of a lot of the internet's activity is generated by bots. So if the massive training datasets contain mostly bot activity, then is the LLM really learning from humans? It sounds like a dystopian game of Chinese Whispers to me. Sure, you can self train your own instance of an LLM to an extent, but can you really trust that it's not going to say something that damages your business's reputation, or gives a customer a blatantly wrong or even dangerous answer?
At YakaJack, we've made a point of not offering any AI integration in our chat clients. Sure, services like ours could do some really useful things, like providing a pre-set message facility so agents can quickly get the boilerplate part out of the way. But outside of the generic stuff, customers should be getting real support from real people.
Just something to think about.
Generated by JackGPT4.0 (Just kidding!)