Comment by victor106
8 hours ago
>No customer or user wakes up and says, ‘I hope I get to talk to a chat bot or an AI agent today
This is so true. I led the implementation of an AI customer service agent and even though management thinks it’s a great success the metrics tell a totally different story. Our customers hated it. I haven’t seen anything in tech that is hated more.
Before you think we did a bad job with our solution, I can tell you we went with some of the best and did our own intensive testing and worked on latencies etc., I actually thought the final version was pretty good but our customers just hated it.
What is is about management where they can't see how bad and half-baked these customer service agents are, but the customer can tell INSTANTLY they're AI and just not helpful in the slightest
They don't use their own product, and they don't want their engineers to use their product either. They want velocity, and you can't have velocity if you're bogged down by doing end-to-end testing and finding friction and whatnot.
When they don't even want to try the thing, they are allegedly proud of, I would expect them to actually now, that the thing is garbage.
1 reply →
Its being pounded into their head almost daily (and from all directions) that AI is the future and the more AI the better. Their boss, the industry, other managers, heck probably even their children.
So any big AI initiative they are apart of had succeeded before it even starts!
Yep, and there's the FOMO aspect.
Especially in non-tech/non-product companies right now. Look at the age profile of owners and CEOs. Most of them were around for the dotcom era, and eventually they saw that companies that ignored the internet or didn't adapt where completely left behind.
There's a real fear among the C-suite/management that this is the same thing all over again, and if they are not fast enough on the adoption they will lose their business.
To them, AI is an existential crisis.
2 replies →
>where they can't see how bad and half-baked these customer service agents are
They don't care, since the sucker, sorry, the customer, keeps giving them his money
Goodhart's Law will always be a factor, and "the bottom line" is largely the ultimate perverse metric for Goodhart's Law. If doing something everyone hates still has a chance to increase revenue, it's still "winning" to management.
Low quality "spam" tactics still reel in enough fish to be monetarily worth the "backlash" from customers that find it distasteful and or start to lose trust in the company. The "We promise we don't spam people" metrics and "Consumer distrust" metrics don't talk enough to the "revenue" metrics, but especially have very different cycles: big customer satisfaction metrics like J. D. Power are big annual things, not quarterly reports like earnings. In my experience things like "how often are we calling the same disinterested people to the point where it starts to feel like very personal spam" metrics don't ever really get prioritized in internal reports unless there's enforcement from Legal departments, and even then Legal departments can't "upset the bottom line" and only care about such compliance when it becomes News and/or Lawsuits, both of which don't even merit even an annual review cycle. (In fact, the modern class action lawsuit pretty effectively prevents that feedback mechanism from cycling, because generally the terms of agreement in a settled class action lawsuit is that the class is no longer allowed to sue again for the same problem, even if the same problem keeps happening and is never actually fixed.)
Unless quarterly earnings reports need to include things like client satisfaction and spam tracking, the only metric management will continue to care about, because it also is the only metric shareholders claim to ever care about, is the "bottom line", no matter how ugly the sausage is made to bump it from quarter to quarter.
Fundamentally, you can convert reputation into current earnings. The former is hard to measure, the latter is much easier to measure. Thus there is a huge incentive for managers to do such. Better run companies tend to understand this problem and are reasonably careful about avoiding it--but when something comes along and upsets the apple cart like AI does they fail to recognize what they are doing.
The incentive to replace employees with ai systems?
VS. the inadequacy of ai systems (nondeterministic output, no reference with reality, unknown signal to noise ratio, low effort etc).
They don't need to see the problem if their competitors are outsourcing customer service to AI as much as they do. It's like self checkout. Businesses love it because it saves them from employing humans at first world salaries.
They care more about short term cost reduction than long term effects of upsetting customers. Even moreso if there is a lack of significant competition, as is often the case for companies doing this, since customers can't just easily switch to something else.
I think it is not cost reduction. AI is not cheap. I think it is mimetic and FOMO driven, they think the press release packed with em dashes will 10x their stock price or something.
1 reply →
If you can save millions in costs every quarter for the degraded experience, I can see why leaders want to take the risk. Really depends on who is making the calls I guess, for the services that depend on customer experience I would think an AI service agent would probably be out of the question.
They know, because in their personal capacity, they hate interacting with AI customer service agents from other companies.
But their bonus depends on driving down costs in their company.
"They know, because in their personal capacity, they hate interacting with AI customer service agents from other companies."
I believe at a certain level, you have real assistants for that.
In my experience, AI customer service agents can be really helpful for some things, but when they aren’t helpful they are the most annoying thing in the world.
I think this leads to a problem because leadership might see metrics showing that the AI service agent successfully helped with 80% of the questions it is asked, but they don’t realize how damaging that 20% is. Over time, more and more of your customer base is going to hit that 20%, so eventually everyone is pissed off at you.
It would be like if 1 in 100 Cheerios in a box were made of poop. It doesn’t matter that most of them are fine, people are going to remember that one cheerio more than the 99 others.
My analogy for this sort of thing is always sand in the lettuce. It's amazing how easily you can ruin lettuce if it just occasionally crunches against your teeth in that particular way: you start dreading the sensation. You can't see it coming, it's not constant, but it's horrible.
I recommend everyone in this thread call OpenAI and talk that support agent. I had some issue and tried to trip it up, spoke naturally and ambiguously, and it actually did a pretty good job.
I feel like this is an example of Sturgeon's Law - almost every thread can be filled with complaints, because most everything is crap.
What's more interesting is the cases where it isn't. Those prove that the idea can be good, but it's obviously a lot more work than "have an LLM answer customer questions".
Money. Often employees are the largest cost center.
The same people who liked outsourcing customer service to cheaper countries are of course going to love doing the same to AI
> the customer can tell INSTANTLY they're AI
I had to call Geico's claims department yesterday and their AI customer service agent had a surfer bro accent, said "no worries" and "hang tight" and "I gotchu" while trying to follow-up on my claim which made it even more infuriating to interact with. Like... stop with the affectations and fake trying to be hip just be professional and succinct please!
It's like how these automated call handlers that say "Just tell us what you need help with" and then no matter how you phrase the sentence it just doesn't understand what you mean. Great idea in theory, horrendous execution. Bonus points if it automatically disconnects after 3 attempts without finding the right magic word from its dictionary of options.
The answer I always give to "tell us what you need help with" is "representative" or "speak to a human". I despise the automated systems that respond to that with "before we can route you to a human, we need to know what you're calling about". Fortunately, some of those will take the continued statement of "representative" as a sign that they should give up and give you a human to talk to. Not all of them, though.
7 replies →
Oh fucking hell yes, how much I hate that shit. Bring back the "press 5 for XYZ" phone trees, I never have problems getting the systems to understand what I'm trying to indicate there.
9 replies →
Really? The modern ones I’ve run into lately are like 100x better than the old flows. “Press 1 for sales. Press 2 to know our hours. Press 3 if you have a ticket. Press 4 to hear our website address…”.
Of course nothing beats a human with real agency at the company but like, these modern agents could be 100x better than what airlines and internet service providers currently have.
They don't care. They have metrics to meet and don't care how. Full stop.
Have you seen "Severance"? It's a wonderful show that shows us a sick truth:
Many people are a different person at work. That different person is devoid of morals & ethics. They are machines intent on meeting metric targets. Nothing else, not managers, not employees, not people, just machines hitting targets they didn't decide on.
How could we possibly expect more?
It’s when they see the money they saved that quarter with layoffs. Then the next quarter customer satisfaction is down and they just don’t know why…
This generalization is false. As a customer I've seen good implementation of LLMs in products and liked it. I don't think they weŕe revolutionary but they filled gaps in the interface. Also the models, so fair, are trained to be useful, honest and to follow user instructions. Which is a breeze from the standard pattenrs.
You may think that it would be better to talk with a human than with a chatbot but the sad reality is that the humans working at customer support behave more like a bot than the LLMs do. YMMV but in most cases I'd rather talk to an LLM.
This reminds me that I have to write my dentist. They replaced their beep answering machine with an AI chatbot, and the experience is horrible. I just want to say what I want, have it transcribed to text, and then have a human do something about it. It. I don't want to have to slowly explain to a bot who is just going to do the same thing.
Plus, the first time you encounter it, it doesn't identify itself as a bot for a couple sentences. And it's convincing enough that you fall for it. The feeling of being let down and realizing that you were just talking to a chump robot is severe, and is now associated with my dentist's brand.
Same experience calling a dental office recently. The voice on the other side introduced itself by name and had this uncanny valley quality where I wasn't quite 100% sure it was a bot and felt weird asking it outright. Made for an awkward conversation. Once it became clear that I was, in fact, talking to AI, I quickly wrapped it up but came away feeling quite negative about the experience. It's not even that it gave bad responses, but pretending to be a human is a step too far.
Same thing happened to me at my dentist! Infuriating!
I suspect someone is selling these to dentists in particular. Dentists have cash to burn on these kinds of solutions, I guess.
I suspect it's more accurate to say the private equity that bought out dental practices everywhere have the cash to burn. At least for now.
2 replies →
I recently took my car to the dealership to get my A/C fixed. On the drive home, I discovered that they'd somehow screwed up my blinkers when they were fixing their A/C. When I called the next day to explain the problem, confirm they knew it was their issue to fix at no cost and to schedule a time to come in I got an AI Assistant. An AI Assistant that has not been trained to expect an issue to be their fault and who kept trying to direct me book on their site or to book a new appointment to get the problem fixed at my cost through her.
The dealership's decision to hand things over to AI, and to choose to focus that AI on only booking appointments instead of fixing problems is a proverbial F U to me. It's the dealership shifting more work on to me. It's disrespectful and wasteful of my time. By the time I managed to get to a human I was angry and distrustful.
The ever dreaded automated phone systems are more tolerable than the AI driven phone systems. The press 1 for... never tried to make you think someone was actually listening to you, yet the AI services are made to come across like you are talking to a human. Don't try to make people think it's a human.
I think the only place so far where I have actually preferred the AI version was when my phone provider switched from a "dumb" chat bot that ostensibly used natural language but rarely parsed anything successfully.
> Don't try to make people think it's a human.
Agreed.
A local pizza place (Tribute Pizza) switched its phone over to an AI assistant that goes out of its way to appear human to such a degree that it inserts random "restaurant bustle" sound effects into the call so it sounds like you're talking to someone standing in a crowded restaurant.
The subterfuge of layering in sound effects to make idiots think they were still talking to a human was a bridge too far and I swore off ever giving my business to them again.
Considering the many embarrassing failures huge fast food chains have had trying to get AI order takers in their drive-throughs I'd be very concerned about an AI at some random local pizza place getting my order accurately
with the keyboard typing and call center chatter... give me a break
I really want to agree and I can fill the rage building inside me when I talk to one... but on the flip side I just had a conversation with the Amazon one and it fixed my weird incorrect region/country problem in about 5 minutes. I was filled with rage the entire time, but it fixed my problem.
I've recently come to the thought that the reason why I find AI so snake-oily is because it isn't solving the bottlenecks of the use-cases it is being applied to.
The problem with customer service was never the frontline support agents but rather that these frontline agents are not empowered to solve the problems they encounter. I once had a human agent admit to me I was wrongly charged but they could not refund me on the spot because of protocol. Replace that agent with the smartest model and it still wouldn't have improved that interaction! (Of course, the business saves money if the AI costs a fraction of a human agent's salary.)
I'll take a shot in the dark and bet there was always an obscure/poorly documented way to solve your problem and that the AI could just find it in its playbook faster; it's search after all! It's also not inconceivable to go as far as to wager that a _human agent_ would just have been as effective; maybe the protocol to do it wasn't some obscure procedure for customer support.
----
That said, this is why I'm in disbelief that AI is bringing in as much value to the table as claimed. I realized that in software, it was never shipping the code that was the bottleneck to profit. I could be the mythical 10x productive engineer but all my output is still gonna be gated by things like testing of all sorts, customer acquisition, product development and design. Testing and product dev you could maybe automate but only after putting in considerable legwork yourself.
And of course, shipping 10x more features does not mean you'll get 10x more profit, not even that you'll get 10x more customers.
I have a friend working for an international law firm which has recently made a big push for responsible AI use. (I won't say which firm but the first partner name has to do with croissants and they recently organized an internal AI congress in Spain.) So, they are paying for AI subscriptions but I sincerely wonder if that's adding to their bottom line since their profit is bottlenecked by (a) how much billable hours they can account for and (b) the judicial process which famously moves at a glacial pace.
It's just a pattern I repeatedly notice when I look closer into things. And of course, as we all know, the cost of AI services today is still heavily subsidized by VC money. When that money is gone, I worry we'll be stuck with _expensive_ AI-centric workflows that's not really adding value to the business.
it really depends on if it can do useful actions, and both you and the company can reasonably trust that it will actually do the right thing
IMO it also depends on having a working alternative for when the fuzzy automated system fails to work... except that's the stuff companies want to eliminate in order to (theoretically) save money.
I went through the text logs for ours and it was all:
"Fuck off and pass me to a human" and stuff like that.
No one wants this who's calling you. We are literally damaging our company with it.
I’ve had to use one of these to book something with a service company. It was horrible. It’s like talking to Pinnochio… there’s nothing there. And it’s trying to sound human and conversational. It’s creepy and annoying.
Just give me a human being or a plain voice menu.
Businesses just don’t want to pay people if they can help it. Some things are inefficient. Get over it.
I was helping an 80+ year old with a phone problem this week. Dealing with the AI CSA was so frustrating. I don't think this lady was ready to burn down data centers that morning, but I think she was looking for a pitchfork by the afternoon.
It depends on the product/service, but if it mostly works and there is a reliable way the agent can escalate to a real human if necessary, I can see potential in many areas. Most support requests did not try out the standard solutions or read any FAQ. So real humans aren't bored to death giving the same answers hundred times a day, but can focus on the real support cases requiring human intervention.
I did actually experience some chat bots that were helpful without needing a human to get my answer quick and some gave a number to call after exhausting their scope. But I am aware, that this is not standard.
I don't immediately hate talking to an chatbot, my dislike comes from chatbots that sound like humans but can't actually do anything useful. Maybe it just searches the KB and returns an article or maybe it just collects information for the ticket and I still need to wait for a human to reply back. Like I know how to search the KB, I wouldn't be trying to talk to support if I didn't have a real issue. Or it takes forever giving my information to a seemingly real person only to be told to wait, just let me fill out a form and give me a queue ETA. And half the time the real agent can't see the AI messages so I'm stuck repeating myself. Or my question is a very simple one that an AI agent could have the capability of figuring out (is the system down? Unlock my account, etc) but it can't because it's incapable of doing anything other than maybe searching the KB.
It's like you hired the dumbest idiot to answer support phone calls who can't do anything other than cold transferring me to someone that can do stuff. It seriously makes me reconsider doing business with them. What critical expense did they neglect in order to have the funds for an AI idiot? It just feels like corner cutting or something to impress stakeholders that doesn't actually improve product quality or worker efficiency, it's just a vanity project.
If you're not willing to build an AI agent that can actually do real work while also being hardened against exploits (see Instagram), do not spend the money. You are either just going to waste money on the dumbest receptionist money can buy or you are giving away the keys to the kingdom to anyone with a clever prompt.
My main issue with these systems is that they tend to not be able to do anything beyond what I can do myself in whatever self-service portal is available. If I'm calling, it's because I need support beyond what the bot can probably provide, and it just becomes a matter of arguing with the robot to get to a human who can actually solve my problem.
> metrics tell a totally different story. Our customers hated it.
Customers hating it only matters to decisions makers if it can causally be connected to lower profit in the next quarter. Otherwise, the immediate cost savings take precedence.
But what if the difference is either talking to an AI now, and the chances are fair that it can resolve my problem vs. staying on the line for a real person, who may be better able to help me, but there is a 2 hours wait (not an exaggeration, Hawaiian Airlines made me wait this long). Suddenly the former doesn't look that bad?
The possibility of being useful sometimes does not imply that putting a button to invoke an AI in every app is actually a good idea.
You reference a customer support line. OK, but would you make it so every use of Google Search starts off with a call to a customer support line before the search results?
That's a circular concern. The only reason there would be a 2-hour wait, would be because the company didn't hire as many customer support agents, precisely because they want people to use the automated systems instead.
If an AI customer service agent disclosed what capabilities it has that go beyond what I can do by myself, that would be an immediate improvement.
People say jobs with the "human touch" will stay relevant after AGI. And I'm like have you seen customer service? I can't even find the phone number anymore on amazon
Whats implied there is that the "human touch" will become a luxury.
I have a hard time seeing a future where retail doesn't bifurcate even further into ultra low margin, happy-path optimized megastores and concierge-style high touch boutiques. Places like Crutchfield that split the difference nicely seem to be a dying breed.
I immediately skip all youtube ads that are AI generated, and I can easily spot them. Not just that, I am planning to actively avoid those brands that are doing this. A small way to send a message that this form of AI sucks. In addition, I make it a point to compliment humans that assist me in customer support, and ask them to tell their management that they need to hire more humans.
The problem is that your principled opposition to these ads by voting with your wallet does nothing. The advantage to the advertiser is that AI allows them to vomit out ads even more cheaply and to make even more micro-targeted versions of them, and the greater number of suckers sold to more than offsets principled opposers.
The real solution would be the ability to block AI-generated adverts completely in your browser, forcing advertisers to send you non-AI ads if they want your attention.
> The real solution would be the ability to block AI-generated adverts
To speak from a cynical step further along the spectrum (e.g. "aacktually the real real problem is...") I submit that "voting with your eyeballs" is just a weaker form of "voting with your wallet", which is itself often a behavioral trap [0] that isn't effective and mainly exists to dis-empower the public.
In other words, blocking the adverts would be nice, but unless it's part of an organized boycott it probably won't affect much either.
[0] https://pluralistic.net/2026/05/21/purity-culture/
The real solution is blocking all ad tech. I'll show anyone who asks me how to set up an ad blocker in a decent browser.
The real solution is to stop using YouTube.
It depends. Sometimes I have really stupid questions and I don't want to read the website, when I know I don't annoy a human when I click "chat" I may use it to ask my stupid question.
But this is a exception most of the time I just try to confuse the AI to get to an actual human faster.
If I were a manager I would be excited as well. Product quality doesn’t seem to be a metric that is actually correlated with executive bonuses, reducing cost is.
It’s why enshitification is so common. Create a tool that quantifies quality in a usable way as a metric and you change the entire economy.
Part of the problem is that even if you did a good job it doesn't really matter because the rest of the industry isn't, so no user wants to give you a chance
How did you handle escalation to a human flow?
Do the customers know they're interacting with AI? If so, is that the intention?
Shitty companies usually outsource to Indian or Filipino call centers, but the AI is a chirpy fake white person.
> even though management thinks it’s a great success the metrics tell a totally different story.
::2065, in a US Social Studies class::
And that, children, across many industries, along with the unfortunately-timed chatbot craze, then believed to be real AI, is the surprisingly simple origin of Corporate Optimistic Cynical Braindeadism. It’s a bit wordy, but it was LLM-generated before the Big Correction, and nobody bothered to fix it.
It wasn't obvious from the project outset that everyone would hate it, no matter how well designed it is?
That’s because customer support should be the absolute last place that an AI agent is used that might be one of the few places you will need humans in the loop for a while because that’s where people go when everything else doesn’t work
>Before you think we did a bad job with our solution, I can tell you we went with some of the best and did our own intensive testing and worked on latencies etc., I actually thought the final version was pretty good but our customers just hated it.
I bet you did a bad job with your solution.
I hate AI (or humans following scripts like robots) customer service because they don't actually provide service. They jerk you around in circles, don't understand basic things, can't help, and take forever.
People don't hate customer service when they feel like they have been served.
[dead]
Guarantee this is a generational split.
The younger demographics will prefer the AI bot to talk to.
Gen Z hates AI… according to https://www.theverge.com/ai-artificial-intelligence/920401/g... (and others)
If that were true schools wouldn't have a problem with AI.
AI usage is rampant.
10 replies →
Yes, it's a generational split.
People in their 60s or older get confused. People in their 40s or 50s tolerate it better. Younger people hate it with a passion and will hate anything that it touches.
The hate Gen Z has for AI products and companies is intense. And they will look at you like a lesser person for using any AI product.
I don't thinks so because the ai bot will reliably give you the answers you could already get from the website to begin with and will never solve your problem. If people are calling or opening an interactive chat, this is because all automatized procedures have already failed and you are in a situation not supported by them.
I think people clicking through websites will be viewed the same as people going to the library and reading through books to research.
You just get the information you need way quicker.
Recently I had to make changes to cancel my flight. Luckily the website had an agent and I used it to cancel my flight. Didn't have to wait for an email/chat or worse call.
I even rescheduled my flight using the same website agent.
It's just way more convenient.
2 replies →
The problem is that the agents can not be trusted to do things so at the end of the day you wade through loads of crap and they cant solve your problem because they usually only have the same powers you do.
Great if grandma doesnt know how to use a web form, fucking useless for everyone else.