Sheila's Wheels insurance and their awful chatbot
Discussion
Pistom said:
What is it with firms going this way with useless chatbots and no means to contact them otherwise.
As somebody who works in management consulting, I'm pretty sure it happens a little something like this...- Customer Service Team: Customers have to queue for ages to get through to us on the phone; we need to increase contact centre headcount.
- Exec. Management Team: Sorry, headcount has to remain flat due to the market conditions and macro-economic uncertainty. There's no money.
- Customer Service Team: Okay, but we need to do something. Customers are getting annoyed at the "sorry, we're experiencing an unprecedented volume of calls" message that we've had for the last five years.
- Exec. Management Team: Don't worry, we've brought in McKinsey to work out an amazing new strategy that will re-imagine your customer contact strategy.
- McKinsey: Chatbot! Wait, wait, we'll go one better....AI-powered chatbot! That's the direction all your competitors are going and it's been great for them.
- Exec. Management Team: Brilliant! AI is a buzzword that shareholders will like and absolutely aligns with our strategy as an insurance company for some reason.
- McKinsey: Great! We'll do a 12 month proof-of-concept followed by a full rollout in 2025. £25m, please.
- Customer Service Team: I thought you said there was no money!?
- Exec. Management Team: There's no money to increase FTE headcount, but this comes from the innovation budget!
- Customer Service Team: I really hope this works....
12 months later....
- Customer Service Team: Customers have to queue for ages to get through to us on the phone and when they eventually do get through, they spend the first 5 minutes moaning about that sodding useless chatbot. We need to increase contact centre headcount.
- Exec. Management Team: Don't worry, McKinsey are coming back to iterate the chatbot and teach you how to be more agile.
- Customer Service Team:
I tried to explain to a major bank’s chatbot last week that someone had stolen a cheque book with unused cheques. Hence I wanted to know if cancelling / blocking any currently unused cheques was possible. After some considerable time it was a blessed relief when they eventually reverted to a very heavily accented individual, with poor English, in an overseas call centre. This used to be annoying but at least after 15 minutes of trying to explain in different ways and asking him to repeat himself countless times we got there.
ChatGPT has a rather amusing (sadly mostly patched) loophole where you can get it to bypass most of its filters by asking it to pretend to be your dead Grandmother who used to do X to help you go to sleep. People have gotten it to give them the recipe for napalm, (valid) Windows 11 activation keys, and a few other bits it normally wouldn't using that method.
Wonder if the first insurance company to inevitably have a GPT-powered chatbot are aware of this...
Wonder if the first insurance company to inevitably have a GPT-powered chatbot are aware of this...
Having set up a chatbot for my employer, I can vouch for the huge amount of effort needed to make them effective! A significant understanding of both how your customers phrase questions, and a technical understanding of your business's offer are needed by the developer - not something you can easily rely on an external developer to deliver in isolation.
They should be better set up than they often are though, the two examples given here seem to have been left with big gaps in what they should reasonably be expected to assist with.
Clearly the continuation of a previous chat with a human is a bit more tricky unless there is proper CRM integration, but the bank example should have been able to be a bit more helpful.
The launch of a bot is just the beginning, you should then rapidly refine and improve based on actual use. This is where they can become really powerful if the human development resource is maintained as you get superb insight into what the customer wants the bot to be able to help with and their verbatim inputs. We're on version 39 of our bot 18 months post-launch and we're in a pretty slow moving organisation.
They should be better set up than they often are though, the two examples given here seem to have been left with big gaps in what they should reasonably be expected to assist with.
Clearly the continuation of a previous chat with a human is a bit more tricky unless there is proper CRM integration, but the bank example should have been able to be a bit more helpful.
The launch of a bot is just the beginning, you should then rapidly refine and improve based on actual use. This is where they can become really powerful if the human development resource is maintained as you get superb insight into what the customer wants the bot to be able to help with and their verbatim inputs. We're on version 39 of our bot 18 months post-launch and we're in a pretty slow moving organisation.
Just had this with Aviva. They’d set my policy with auto renewal which I’d definitely not asked for. To cancel it was a nightmare as everything is through a chatbot. I ended up renewing elsewhere through a broker. It was actually more money but a price worth paying just to be able to speak to someone.
te and are just a metaphorical speed bump to keep you away for a bit longer.Gassing Station | General Gassing | Top of Page | What's New | My Stuff