The author argues that customers do not actually want chat bots for customer service, contrary to what companies claim. Chat bots can only handle simple, routine queries, but for complicated issues customers want to speak to a human representative. Companies are pushing chat bots to reduce costs and increase profits, without considering the negative impact on customer experience. The author only sees chat bots as useful for customers when used to cancel subscriptions that require contacting customer service, showing how frustrating the current system is. The author believes we should build technology that customers actually want and would appreciate, rather than focusing on bad experiences or defending against them.
Customers want their issues to get solved… but that ship has sailed a long time ago: first tier support, is often outsourced to call centers which are given a very strict list of subjects and procedures to follow; if a customer’s case is not in there, then they’re SOL.
What’s worse: call center companies, accept contracts from multiple companies that want to offer support, meaning the people working at a call center now have to learn not just one company’s script and strict guidelines, but those of multiple companies at once.
If we add the fact that these call center companies pay peanuts and have poor worker retention, there is close to zero chance a customer will contact a first tier support worker who knows all the strict guidelines they’re required to follow from the company the customer is seeking support for.
Chat bots are not a general solution to all customer support, despite their overhyped marketing, but they are a solution for “first tier agent knowing each and every strict guideline by heart”. Now each company just needs to feed their predefined procedures to an AI, and customers will never again call someone who has barely any clue and needs to fumble around for half an hour just to give a wrong answer.
From a consumer’s point of view, it’s like having access to a 100% accurate search engine into the company’s predefined procedures… which might not sound like much, but is still better than the current state of affairs. For anything not prepared ahead of time in the company’s support book, customers will still need to ask to escalate as usual… or even get escalated transparently when the bot realizes it can’t provide an answer.
A few of the chatbots I worked on, back when I did that, were actually good. Those companies had actually looked at their support traffic and figured out that like 95% of it was people asking the same 20 or so questions that had specific answers. Or at least that you could get to a specific answer with 1-2 followup questions. Like, a huge number of people just want to know how to pay their bill, and the answer is “go to this webpage or call this number”.
It’s kind of a waste of human time and effort to have a human answering all those questions, so the chatbot dealt with those (and tbh it was 50-50 whether those people even knew they were talking to a robot) and the actual hard shit got a warm transfer to a human agent who got the chat transcript.
Honestly the companies it worked best for, either their online documentation was a total shitshow so the chatbot was your best hope of actually finding anything, or a huge proportion of their customer base were total luddites who just didn’t want to use a website and wanted to talk to someone. (We had to make our chatbots support Internet Explorer 11. In 2021. Because for some of our clients IE11 was like 30% of their traffic. I don’t even fucking know.)
Tangential, but my last employer (US based) outsourced L1 IT to a call center in India, and it was maddening. They didn’t know very much beyond the script, and often you just had to say the right words to get your issue escalated, but it would always take a day or so to get called back. It drove me nuts as an engineer, but I’m sure it works fine for people who are less familiar with computers.
I used to design and maintain chatbots for a living, for a company that among other things sold bespoke chatbots to corporate clients, and I can tell you that the companies KNOW that customers don’t want chatbots for customer service. They don’t care. THEY want chatbots for customer service because chatbots are orders of magnitude cheaper than hiring customer service representatives.
A chatbot is gonna cost what it costs them to employ 1-2 customer service reps, but it can handle basically infinite traffic for that price. The GOOD ones handle the simple questions (your "how do I pay my bill"s and your "what are your hours"s) and then forward the difficult ones (“why is my bill fucked up?”) to a human agent. But I absolutely worked with some clients (who I will not name because I do not want to get sued) that explicitly wanted to avoid letting customers get access to a human agent by whatever means possible.
Also a side note but basically no one lets people cancel accounts via chatbot. They inevitably want THOSE requests to go to a human rep so they can try to talk them out of it.
It there is an issue you have that you can’t figure out from the website but the chatbot is capable of solving you should make a better website.
Half the “support chat bots” I’ve talked to is just a paraphrased version of searching their support article database. If it’s not in there I pretty much have to talk to a real agent.
That said I don’t think companies would want chatbots that could do more than that, at least for the time being.
They could end up being convinced into giving me an 80% VIP discount without the company’s consent.
E: fixed a they’re i was tired this morning
When I contact customer service I almost never want information, I want them to do something. As long as the bots can’t actually make anything happening, they are just a waste of my time. And that’s why I don’t like them
Most chatbots are speed bumps. Like phone menu trees and hold times, they slow you down on your way to get actual help.
Sometimes that means you give up before getting to the real help, which saves money on support.
Whether it’s the intended effect or not, it is so well known at this point that we shouldn’t excuse anyone using this tactic. It’s malicious.
I agree that I absolutely do not want a chat bot and that getting in touch with a human who knows what theyre doing is more important.
That said I’ve also worked a job where I manned the little chat box on a website by myself and I was consecutively talking to like 13 people at the same time for 8 hours. It was not fun, and while I did what I could to help people there were times I wasnt as fast to respond and that I didnt give people as much help as I could. There were also times when the the question was super simple and it saved the customer time on hold for nothing.
I’ve also seen chat windows on websites that are pushed out to underpaid overworked people in third world countries where they are so stuck to the script that it might as well be a robot. Overall I think chat windows on websites for anything serious arent great, human or otherwise, though they should be better. In some cases the bots may improve experience, but I dont like that it’ll just lead to cutting their customer service crew further.
I’ve done chat at my internal IT job, and I’ve been talking to 5 people troubleshooting, with me remoted into 2 of there computers and it is insane
In my country they are thinking about putting chat bot on the emergency line (same as 911 call for reference).
So no…when I call I want help, not a chat bot with limited options, no empathy and that will probably desconect my call if I choose the wrong option.
That would bring fear to my life. I have not had any good chat box experiences and certainly would not want one during a potential emergency. Some countries - perhaps most - have a dedicated line for real emergencies and a separate line for non emergency calls. I would be frustrated if the officials put a chatbot on either of them.
Wtf, which country are you in?
Portugal
https://24.sapo.pt/atualidade/artigos/tecnologia-do-chatgpt-vai-gerir-atendimento-de-chamadas-do-112-em-2025 It is in portuguese, but I think google can translate the webpage
I don’t know what type of chatbots these companies are using, but I’ve literally never had a good experience with them and it doesn’t make sense considering how advanced even something like OpenOrca 13B is (GPT-3.5 level) which can run on a single graphics card in some company server room. Most of the ones I’ve talked to are from some random AI startup that have cookie cutter preprogrammed text responses that feel less like LLMs and more like a flow chart and a rudimentary classifier to select an appropriate response. We have LLMs that can do the more complex human tasks of figuring out problems and suggesting solutions and that can query a company database to respond correctly, but we don’t use them.
We have LLMs that can do the more complex human tasks of figuring out problems and suggesting solutions and that can query a company database to respond correctly, but we don’t use them.
I don’t think we have this at all. We have something that can sometimes appear to be that, but falls well short.
I would prefer a documentation site with a fuzzy finder, where I can search terms and the articles are well written, and If I don’t find my answer I would like to contact a real person. Chatbots are very inconvenient for finding information, and they are also slow. Maybe something like this https://support.system76.com/, https://wiki.archlinux.org/, https://wiki.archlinux.org/. good docs save more time that those crappy chatbots, and a way to have a cal with a real human. (Maybe chatbots if they were something like chat gpt)
I have never interacted with one of those chat bots that didn’t lead to me just speaking to a representative anyway. Why the extra steps.
I think the worst application of chat bots is when they replace a form that is served on a webpage. I don’t know why anyone thinks this is a good idea but I’ve seen it a lot.
deleted by creator
If I had a simple, routine problem, I wouldn’t be calling you; I’d be solving it myself.
I think there are ways they could be used. If they’re intelligently used as a filter to resolve basic issues that don’t require human intervention and immediately transfer to a human when the query doesn’t fit into a bucket extremely cleanly, they could be an efficient force maximizer, and you could theoretically even have specialists in different areas of your service that the bot attempts to direct most relevant queries to.
But the companies using them don’t want great, efficient service. They want cheap service at all costs to anything else. So that’s what you get.
