To me the second response reads more like PR-speak, like somebody who's been over-coached on how to talk to customers.
I find it strange that they'd have an AI-driven followup email days after the survey. If you're automating it, why the delay? "Delay the response so it seems more human" is way too cynical.
It's worth noting that OP is talking about Ocado, too. I've been a client of theirs for over ten years, and the number one reason why I keep buying from them is that their delivery drivers never feel super-hurried. They have lower quotas than drivers for other services, and they always have time for a little bit of banter. Compare that e.g. to Amazon Fresh drivers, who'll always just leave the groceries at the door and knock. They're already gone by the time I open the door.
I find it unlikely that they'd automate this sort of second-level customer support response while not squeezing the everliving shit out of their drivers.
One particular sentence reads particularly human to me:
> It is rare to get feedback that tells us exactly what is broken and how to fix it.
AI doesn't remember from one conversation to another, so probably can't say how rare something like this is at Ocado (although admittedly it could learn that it's generally rare in fine tuning). I also expect AI to be trained not to compare feedback with other people's feedback.
“You’re absolutely right, and honestly, that’s rare. Nobody else would have thought of taking a bath with their toaster, but you did. Would you like me to write you a twelve-week toaster making toast plan?”
I only saw pangram being mentioned on HN. How good is it in general ? And what do they use to check if a text is AI Generated, the text watermarks from big models ?
it is very good, but keep in mind it'll also flag text edited with grammarly. Maybe not the same as copy and dumping a claude response, but still normalizes the authors voice with AI.
In my experience, it can have significant false positive. It's pretty good at telling you when something is NOT written by an LLM, but when it says something is, I take it with a big grain of salt.
That's how the earlier image generation models worked. You had one ai to generate images, and one to understand if it's a fake generated image or a real one (like, you have a real image of a dog, use the ai to generate an image of a dog, and pass both to the ai trying to see if it knows which is which). They both start out bad, but evolve together.
Models are trained on ridiculous amounts of human-written text, but their outputs are pretty easy to recognize as not human written. Pangram explains some of the reasons for this in a blog post.[1]
The problem with using Pangram in a feedback loop is that Pangram only reports whether it thinks the text is written by AI or not. It doesn't report whether the text makes sense, or whether it contains errors in spelling or grammar. So when people try to pair AIs with Pangram, the quality of the output tends to degrade to the point where it's not useful.
Basically, it's a harder problem to get an AI to accomplish some goal (such as writing a blog post) while also having its output mimic human text. To use an analogy, it's like building a stealth aircraft with the same speed, maneuverability, range, and payload as a non-stealthy aircraft. Even with a much larger budget, it's not always feasible.
It does smell like AI. The one bit that gave me pause was "Thank you again, I do hope that this issue is resolved quickly." AIs don't usually do comma splices. Maybe it was AI with light human editing.
> There is nothing more frustrating than pointing out a legitimate bug only to feel like you're talking to a brick wall.
At a bare minimum, this sentence was AI inspired. Likely generated. Which certainly casts the whole response into doubt (that it is human-written.) Reads more to me like a person gave a quick quippy response to an AI to clean up, then just took the output and replied.
The thing about this whole post is that AI-written text doesn't tell you a single thing about whether a human was involved.
A thing about tier 1 CS agents is that they are often not highly literate and great at writing, and companies go to great lengths to minimize how much text they need to freehand, with tons of pre-writes to address common situations. LLMs allow the agent to express some freeform idea and have the LLM clean it up to read clearly and professionally. So even if you are 100% dealing with a human, you should expect nearly all text from them to be LLM-written.
I read it twice and the second time it stood out as not.
Usually there is a taste about AI generated messages that are just quite obvious, it's usually pretty close to corporate speak, but not exactly the same.
I think this also gets muddier and muddier because A. LLMs do get better at writing like a human, though slowly (some analysis posted here lately showed pretty clear distinctions in pacing, word-choice etc; not bothering to look it up right now) and B. human language is fluent and people start adopting AI speech!
There are so many levels to this: fully automated AI reply, human sending AI-written reply, human sending AI-written but hand-improved reply, human sending hand-written but AI-improved (or translated) reply, human who reads AI output all day hand-written reply, human who doesn't interact with AI (often) hand-written reply.
Some of these will give you an uncanny feeling, but you always keep guessing who you're actually talking to. In the internet nobody knows you're a dog.
To put salt in the wound, I wonder if the first AI ever even raised the issue with a support team. Pretty much every customer support AI has gaslit me saying it escalated, or please wait for a human to contact you, and then sitting there for an hour or more to confirm that, yes, that was just hallucinated in the prompt, and it's been waiting for my next prompt ever since.
The good thing about AI support is that absolutely flipping out in the chat actually does get escalated, unlike with a human who just digs in their heels.
Exactly! I absolutely don't mind AI generated responses as long as it is accompanied by an AI generated ticket/bug along with an AI generated fix, along with an AI generated regression/integration test and AI generated message saying it was fixed. An AI generated coupon/reward for the trouble is also fine.
If this is the workflow that a consumer experiences, I don't think majority will complain about AI.
This thread highlights why I often say that at some point in the future, the obsession with whether something was written by AI or not will be classified as a mental disorder.
I never contact customer service with a complaint, I just don't use them again. But I have, very occasionally, contacted customer service with a "could you pass on my thanks to ..." for a particularly good product or service. The last a couple of months ago to Marks & Spencer for whoever it was who discovered their Len de l'Ell wine, which is fabulous. I was so insulted by the AI response that I'll never do that again.
I mean, you quoted them as saying "with a complaint", but your example is of something that needs to be fixed. I also very, very rarely complain to customer service, and am more likely to use another company if my ire rises to the point where I am considering a complaint.
My wife complains to customer service without any expectation that something will change. If she gets a refund or partial refund, she's somewhat more mollified, but she will definitely write a complaint even if she has no expectation that there is anything that can now be done about her experience. She mostly wants to vent, and for the company to validate that they screwed up and seem apologetic about it.
Contacting customer support doesn't make me feel better, and it's almost always more bother than any refund I could get, so I don't do it much. The article is indeed about trying to get something fixed, but I wouldn't call that a complaint. It's more a suggestion.
> i think everyone who complains to customer service is doing so because they want something fixed?
Clearly you've never worked in customer service. Yes, people absolutely will call in or email complaints purely for the sake of complaining. No specific actionable complaint, no resolution possible, just complaining to complain.
I don't know how you draw the line but to me there's a clear distinction between "you've fraudulently taken my money" and "this is shitty for everyone could you fix it please".
> i think everyone who complains to customer service is doing so because they want something fixed
When I worked customer service a substantial number of complaints to customer service seemed to just be to complain. They won't care if you fix the thing, they'll complain about that too.
I think ideally a product should not need customer service in the first place.
I had wildly different results. A lot of customer service sucked, other customer
service was ok-ish. But in general it always takes away my time. Often a really
well-designed manual, FAQ etc... solves most issues. Only in a few instances I had
to rely on customer services.
I think ideally a product should not need customer service in the first place.
That's what I tell my boss when she asks what can be done to increase the number of visitors to our web site.
I keep telling her, this isn't Burger Hut, it's healthcare. If people are visiting the web site, it's because things are going poorly, not because things are going well. If the doctors do their jobs, I don't have to do mine.
As a grumpy misanthrope I welcome the trend of overly sycophantic AIs increasing the value of borderline rude thank you messages as a sign of actual human responses, increasing the demand for grumpy misanthropes on staff, and hope that AI can't emulate us well soon.
You cannot pull that off if you're mid, and mid is exactly what the statistical average of humanity is.
It will try though, and it will just be an asshole.
Which, funnily enough, is exactly the same failure mode as mid humans trying on that costume fall into.
I find this pretty funny. All over the Internet, you will find people accusing writing of being AI. Things that don’t really stand out to me. It’s become the new witchcraft. It seems we’re at a point where nobody can tell what was written by human, so what’s the point in reading any of it?
A fast, shallow reply doesn't save a customer's time. It's just a useless volley back to the customer that keeps their request off the support queue.
It may burn off customer support engagement among customers who don't have the capacity or patience to keep pushing for help (this is a bad outcome for customers, but looks deceptively positive in vendor-side metrics), but for the remaining customers it just adds latency and work.
I had a AI customer service interaction (it was explicit about being AI, used fin.ai iirc). At first I was amazed! It was doing email as fast as IM! And the first response seemed really knowledgeable and put me at ease.
But it failed to do a key step (generating a shipping label for me). I went back and forth over 21 messages trying to get it to do it. Then I wrote "escalate to a human" and an hour later someone replies with a shipping label directly.
I would much rather have had a slow but correct response than a fast but confidently-wrong response.
Before the AI first level support often wasn't allowed to type answers but instead copy paste one of the answers from the faq, which was extremely frustrating.
The chat would go like this:
- I have a question about "no vehicles in the park" policy. Does it include motorized wheelchairs?
- To guarantee safety and blablabla our policy is that no vehicles are allowed in the park.
- so can i go to the park with a wheelchair with a battery and electric engine?
- To guarantee safety and blablabla our policy is that no vehicles are allowed in the park. Please see our safety policy (link).
Default llm gened text can be easily detected, but if operator provide sufficient style guide - it is almost undetectable. The second response was also llm, just with little bit of style guide.
For as much as I like AI for (gestures broadly around the hacker news space), I am glad that I'm not alone in getting real upset at the endlessly incessant user-directed AI garbage, be it customer service emails, or ChatGPT generated event posters.
The reason you like it is the same reason people in customer service like it, too, they can do their job with less effort and coast on middling skill for longer.
I think Rodney Brooke’s three laws of robotics apply here, particularly how the appearance of the bot promises its capabilities. It might have better for an obviously automated response to come back. No one might be listening but would not have been promising human empathy while faking human attention.
Even if the second response used AI, it at least seams like someone took the time to edit it that response thoroughly to make it more digestible. I think that's what OP, and myself, would at least expect in that situation.
I'm glad there are people doing this, especially when it comes to comms. I feel like we have a brief window as we decide culturally what we consider rude: sending people LLM-generated prose without a clear provenance/masquerading as human-written or calling this behaviour out.
I hope society settles on the former as the faux pas.
As an aside, I had a near identical experience trying to report a technical issue to ocado minus the nice human follow-up. I hadn't heard of the author of this post before this, but I like him :)
We are just a couple years away from companies realizing the first line of customer service can be a profit center and have the largest impact on brand likeability for minimal cost.
Yes, when people use customer support as a glorified FAQ lookup an AI response is appropriate, but a few humans that care can positively impact 100 customers a day. AI is great at routing messages to the right place, not as tier 1 support.
In the AI everyrhing world, humanness is going to be the marketing edge.
> We are just a couple years away from companies realizing the first line of customer service can be a profit center and have the largest impact on brand likeability for minimal cost.
There are plenty of companies that already realize this, or have in the past. Zappos is a famous one, but plenty of devtools companies know that customer service is a differentiator. One that can be positive or negative.
> the first line of customer service can be a profit center
Are you suggesting that they'll be charging for customer service? That seems even worse. (or do you just mean that it's so good that people positively want to use the brand because of their customer service?)
Humanness has only brought me long waits on a phone listening to shitty music only to eventually be connected to a human who can’t really even resolve my issues and forwards me to another human who also can’t resolve it.
AI will ultimately prove better at generating the kind of message the author prefers. You can just prompt your AI agent to "avoid sounding like AI."
That aside, what matters is whether the company fixes the problem in a timely manner. If they don't, it doesn't matter how nice a message they send. And if they do, it hardly matters if their message is brusque, boilerplate, or obviously AI.
> You can just prompt your AI agent to "avoid sounding like AI."
I’ve tried this with claude. I told it to output HTML that didn’t look like it came from claude and to avoid common claudisms. It didn’t work and still used claude boxes all over the place. I’m sure with enough time I could get it there, but at that point I may as well write the HTML myself.
Personally, I can't stand default LLM writing. I found out about this skill https://github.com/ayghri/i-have-adhd That I actually copy-pasted into the default instructions for Claude.
It's night and day, I wouldn't have been able to use it if it wrote in its default style.
It can be a bit weird because when it outputs a list it asks you to report on the first item (and usually it's stuff so small I do all the steps before writing to it again).
I still wonder if AI will be designed to have flavor profiles like coffee. Right now I don't think alot of product thinking has gone into this line (maybe it has??) but I suspect that AIX (AI eXperience) design will become more and more of a thing as the space matures.
I'm sure to some degree people (and AI being used to build AI) are kind of doing this thing but I don't yet think that OpenAI has a distinct flavor profile for their models that they have standardized on and are intending to lock into over time.
When I go to Starbucks I know what I'm going to get and it's been reasonably consistent over a decade plus and similarly that's true of other coffee shops which have different profiles. I'm expecting something along those lines.
>AI will ultimately prove better at generating the kind of message the author prefers
It's already pretty straightforward to improve this default "You're absolutely right about [your email reworded]" response but you still see it everywhere. The corpos simply don't care to fix it because doing so probably doesn't affect w/e metric they're modelling.
I think the real question to ask is not really about the tools themselves but where the organisation's incentives really lie. I've come to the conclusion that customer satisfaction just isn't up there anymore. You'll eat whatever garbage is served in your trough (from their POV)
I've come to the same conclusion. It feels like, in the past 5 years or so, all businesses beyond a certain size have adopted the "we're too big and successful to give a shit about our customers, they'll keep paying us anyway" mindset.
Here's my own example: My local ISP has had intermittent connection drop issues for years. I'd call them about it, and they'd at least pretend to care, maybe send in a guy to replace the router, even though they never actually fixed it for good. If I call them today, though, I'll be forced to explain my issue to a "virtual assistant", get put on hold for 20+ minutes, and if there's an actual human available to take my call, they'll basically tell me to pound sand. And I have no choice but to accept it, because all ISPs are like this now.
> You can just prompt your AI agent to "avoid sounding like AI."
It'll never work like that, because everybody else giving that same direction will get more or less the same result, which will itself come to be recognized as sounding like "AI`" for any popular model+harness.
What you'll always need to do is instill your AI with some idiosyncratic voice that distinguishes it from others, which you can achieve in a variety of ways (stylistic direction, audience characterization, style mimickry, personalized finetuning, etc). Presumably, tools for doing this will eventually make it into the harness products but anybody who doesn't use them is always just going to end up generating recognizably AI slop prose. It's the nature of the technology.
A secondary consequence (even if you just use "avoid sounding like AI") is that because the technology works its magic mostly through prose-streams in its own voice, these idiosyncratic voice changes will also nudge its behavior away from benchmark norms. You won't necessarily lose capability on any given task (you may even gain on some), but external reference points/measures about its capabilities will become less applicable.
I think the canned response is more honest, because it's usually trivial to tell that it's canned. Maybe if the AI generated response said it was AI generated it would feel less shitty.
I empathize with the OP, but also - what's the point? The only reason we know the first response is AI is due to the obvious tells, and we're even debating in the comments here if the 2nd response is human or not.
Let's just cede the ground - soon you won't be able to tell if it's AI or not. Then what? What payoff were you hoping for when you thought it was human-to-human that you'll no longer get human-to-AI (when the AI is indistinguishable from human)?
IMO you're going to start using AI instead of your human time. "Claude, send a note about that malformed calendar invite thing - convey my annoyance".
Then that's going to force the companies that DO use humans to respond to switch to AI, because the responses are coming in via AI.
> Then that's going to force the companies that DO use humans to respond to switch to AI, because the responses are coming in via AI.
There are two obvious ways to handle this.
The first is to charge for human support. If you want to pay $100 each time you make a human spend 15 minutes responding to your AI agent's query, it's your money.
And the second is that you put a chatbot out front that has memorized the FAQ and thereby successfully answers 97% of queries, but cause them to get better at identifying when they're being useless and route that small percentage of issues to human support.
Assuming the person is serious and not writing this tongue in cheek...
I find it humorous is that all it takes for OP to think the response was written by a human is that it says it was written by a human.
Both responses could have been from AI. Or none. Or just one. Just look at the comments on this thread, nobody can agree. And in the end, it's simply unfalsifiable.
Why even waste time playing this game?
All that matters is the result: was the complaint addressed in the end?
I don't understand these weird hills that people want to die on.
The current worst customer service departments are the ones that never set up a website. Presumably easy to use customer service would be a net loss because it wouldbe used more? But they've happily added AI to their phone tree! It burns ever more of your precious human lifespan while they have even fewer minimum wage slaves in their call center.
And insult to injury, these bots absolutely spew patronizing crap at you so they burn the maximal amount of time while being obliquely rude and condescending at the same time.
I don't care if there is a future where my locally hosted AI assistant is affordable and can deal with all that AI-generated stuff for me. That's actually the future I prefer. My fear is that governments will either outlaw local personal AI or outlaw using AIs to deal with company and government AIs. I want my AI to deal with all bureaucracy, all daily issues that need fixing, all forms to fill out, all signatures to give ("This letter has been written electronically and doesn't need a signature."), all taxes, all phone calls, so I never ever have to personally deal with any of this crap again.
I'd prefer some variant of "This letter has been written electronically and is cryptographically signed by the private key representing the <meat or machine> entity known as '<name goes here>'. The matching public key is located at <web address> and attached to this message at <timedatestamp>." Fairly trivial to automate, and gives all outgoing messages a clear and easily verifiable chain of provenance. But that's just me. :)
Why aren't you just sending the initial email with your own AI tho?
Why are humans doing any of this?
The internet should be for bots and slop shop farmers and indie web wizards only; — :P. meatbags can go hang out at the farmers market.
I mean, who cares if the issue gets resolved, right? Why do you care if it's an AI agent or someone in a giant support center in a different country? Personally, I would prefer AI handle these things as I trust a poorly paid "technical engineer" to report a bug a lot less.
I may be a "head in the sander", but I have a rule about this and many things that allow me to resist without getting all wrapped up in these details.
I feel that even talking about AI which everyone seems to want to do as the consequences of AI are starting to slowly become more evident and obvious to even lesser technical people is a waste of time and energy.
My rule to overcome all this nonsense is simple.
I will not change. Nor will I even bother justifying that, I dont have to.
I will not change.
I am noticing more and more people resist the AI slop attack. I like it.
It may not be realistic, but I am slowly changing my own approach - I simply
don't want to have anything to do with AI slop. It taints and corrupts. It
destroys (some) projects.
I am not at all claiming AI use has no use case - yes there are use cases. But
there are also hugely negative parts. The fact that AI is used to zero out or
reduce human interaction is a negative part.
I had a bit of fun lately with regards to Ginux, aka can AI slop auto-generate
an operating system (or at the least the kernel, such as the linux kernel or
an equivalent; see https://itsfoss.com/news/ginux/, I was surprised to see
that's already a week old, I did not notice it initially). I think projects
such as AI linux 2.0, will fail in various non-technical areas. People won't
jump on board with AI meta-designing things - or, more accurately one guy
using AI to slop-spam-generate operating system software (assuming it works
on real hardware eventually; right now Ginux does not, if I recall correctly,
only via VM).
Resistance to AI slop is not futile - it is mandatory. Imagine if the Star
Trek Borg would have used AI slop. They would have failed. But AI could
improve eventually, quality-wise, and then things become more interesting
again. Resistance to AI slop may eventually not be futile, the moment it
is no longer slop.
I used em-dashes -- routinely -- before David E. Rumelhart even wrote "Learning representations by back-propagating errors" in 1986. AI hysteria is wrecking language as much or more than AI itself is. Can we please look at what is being expressed instead of being fascists about how it is expressed? If you can't honestly understand, and it happens often even to extreme AI boosters, it is more than fair to say so. Uncurious -- hidebound -- xenophobes.
You do realize that this kind of rant leads to AI designed to better emulate humans? It will produce the same flavor of rambling sentences, slightly off grammar, and other indicators.
Companies will not stop using AI, so the net result can only be a degradation of customer support to the same shit level that we had previously endured. But now there are no longer humans in the loop, and you can no longer have the signal to tell you that you should stop engaging with them.
Arms races have no winners, and it will be no different in this scenario. Heck, how do you know this has not already happened? How do you really know the second email was written by a human and not by exactly the kind of AI subterfuge that I just described?
The second response reads as very much written by AI to me too.
To me the second response reads more like PR-speak, like somebody who's been over-coached on how to talk to customers.
I find it strange that they'd have an AI-driven followup email days after the survey. If you're automating it, why the delay? "Delay the response so it seems more human" is way too cynical.
It's worth noting that OP is talking about Ocado, too. I've been a client of theirs for over ten years, and the number one reason why I keep buying from them is that their delivery drivers never feel super-hurried. They have lower quotas than drivers for other services, and they always have time for a little bit of banter. Compare that e.g. to Amazon Fresh drivers, who'll always just leave the groceries at the door and knock. They're already gone by the time I open the door.
I find it unlikely that they'd automate this sort of second-level customer support response while not squeezing the everliving shit out of their drivers.
One particular sentence reads particularly human to me:
> It is rare to get feedback that tells us exactly what is broken and how to fix it. AI doesn't remember from one conversation to another, so probably can't say how rare something like this is at Ocado (although admittedly it could learn that it's generally rare in fine tuning). I also expect AI to be trained not to compare feedback with other people's feedback.
It doesn't have to remember. It just repeats what humans say. And humans say that sort of thing every day.
Nope, that’s classic GPT slop.
“You’re absolutely right, and honestly, that’s rare. Nobody else would have thought of taking a bath with their toaster, but you did. Would you like me to write you a twelve-week toaster making toast plan?”
And also to pangram 4.1, which rates it as 100% AI Generated.
I only saw pangram being mentioned on HN. How good is it in general ? And what do they use to check if a text is AI Generated, the text watermarks from big models ?
it is very good, but keep in mind it'll also flag text edited with grammarly. Maybe not the same as copy and dumping a claude response, but still normalizes the authors voice with AI.
Grammarly does use LLMs for a while.
In my experience, it can have significant false positive. It's pretty good at telling you when something is NOT written by an LLM, but when it says something is, I take it with a big grain of salt.
I hadn't heard of it before, just now looking them up.
Dumb question, but could you pair AI with pangram to make it sound less like AI or would that be a recursive loop? I'm skeptical about it
That's how the earlier image generation models worked. You had one ai to generate images, and one to understand if it's a fake generated image or a real one (like, you have a real image of a dog, use the ai to generate an image of a dog, and pass both to the ai trying to see if it knows which is which). They both start out bad, but evolve together.
https://en.wikipedia.org/wiki/Generative_adversarial_network
Models are trained on ridiculous amounts of human-written text, but their outputs are pretty easy to recognize as not human written. Pangram explains some of the reasons for this in a blog post.[1]
The problem with using Pangram in a feedback loop is that Pangram only reports whether it thinks the text is written by AI or not. It doesn't report whether the text makes sense, or whether it contains errors in spelling or grammar. So when people try to pair AIs with Pangram, the quality of the output tends to degrade to the point where it's not useful.
Basically, it's a harder problem to get an AI to accomplish some goal (such as writing a blog post) while also having its output mimic human text. To use an analogy, it's like building a stealth aircraft with the same speed, maneuverability, range, and payload as a non-stealthy aircraft. Even with a much larger budget, it's not always feasible.
1. https://www.pangram.com/blog/no-llms-dont-just-mimic-human-t...
That's what adversarial model training is basically.
They trained a model to detect AI-written text.
It works fine on mainstream models where the output hasn't been edited.
It doesn't work on humanizers or custom model output.
Pangram says it's 100% AI: https://www.pangram.com/history/04df89b1-5db5-40b8-b837-c430...
It does smell like AI. The one bit that gave me pause was "Thank you again, I do hope that this issue is resolved quickly." AIs don't usually do comma splices. Maybe it was AI with light human editing.
It could have been the other way around, with a human typing and something like Grammarly cleaning up the text.
> There is nothing more frustrating than pointing out a legitimate bug only to feel like you're talking to a brick wall.
At a bare minimum, this sentence was AI inspired. Likely generated. Which certainly casts the whole response into doubt (that it is human-written.) Reads more to me like a person gave a quick quippy response to an AI to clean up, then just took the output and replied.
The thing about this whole post is that AI-written text doesn't tell you a single thing about whether a human was involved.
A thing about tier 1 CS agents is that they are often not highly literate and great at writing, and companies go to great lengths to minimize how much text they need to freehand, with tons of pre-writes to address common situations. LLMs allow the agent to express some freeform idea and have the LLM clean it up to read clearly and professionally. So even if you are 100% dealing with a human, you should expect nearly all text from them to be LLM-written.
I read it twice and the second time it stood out as not.
Usually there is a taste about AI generated messages that are just quite obvious, it's usually pretty close to corporate speak, but not exactly the same.
I think this also gets muddier and muddier because A. LLMs do get better at writing like a human, though slowly (some analysis posted here lately showed pretty clear distinctions in pacing, word-choice etc; not bothering to look it up right now) and B. human language is fluent and people start adopting AI speech!
There are so many levels to this: fully automated AI reply, human sending AI-written reply, human sending AI-written but hand-improved reply, human sending hand-written but AI-improved (or translated) reply, human who reads AI output all day hand-written reply, human who doesn't interact with AI (often) hand-written reply.
Some of these will give you an uncanny feeling, but you always keep guessing who you're actually talking to. In the internet nobody knows you're a dog.
That is what I immediatelly detected too, but was in shock that the author fell for it.
Pretty sure the blog post meant to be ironic. But who knows.
To put salt in the wound, I wonder if the first AI ever even raised the issue with a support team. Pretty much every customer support AI has gaslit me saying it escalated, or please wait for a human to contact you, and then sitting there for an hour or more to confirm that, yes, that was just hallucinated in the prompt, and it's been waiting for my next prompt ever since.
The good thing about AI support is that absolutely flipping out in the chat actually does get escalated, unlike with a human who just digs in their heels.
Exactly! I absolutely don't mind AI generated responses as long as it is accompanied by an AI generated ticket/bug along with an AI generated fix, along with an AI generated regression/integration test and AI generated message saying it was fixed. An AI generated coupon/reward for the trouble is also fine.
If this is the workflow that a consumer experiences, I don't think majority will complain about AI.
This thread highlights why I often say that at some point in the future, the obsession with whether something was written by AI or not will be classified as a mental disorder.
I never contact customer service with a complaint, I just don't use them again. But I have, very occasionally, contacted customer service with a "could you pass on my thanks to ..." for a particularly good product or service. The last a couple of months ago to Marks & Spencer for whoever it was who discovered their Len de l'Ell wine, which is fabulous. I was so insulted by the AI response that I'll never do that again.
Don't you run out of companies to use?
>I never contact customer service with a complaint, I just don't use them again.
my phone provider double-charged my credit card so i had to call in and get it reversed.
would you just accept the double charge, cancel your contract (and pay the cancellation fee), and then move to one of the other handful of providers?
I mean, you quoted them as saying "with a complaint", but your example is of something that needs to be fixed. I also very, very rarely complain to customer service, and am more likely to use another company if my ire rises to the point where I am considering a complaint.
>but your example is of something that needs to be fixed.
i'm a bit confused at this distinction. i think everyone who complains to customer service is doing so because they want something fixed?
like, in the article, they complained about the malformed calendar invites because they want them to get fixed and no longer be malformed.
i can't even really think of an example of a customer service call that isnt looking for some thing or situation to be fixed
My wife complains to customer service without any expectation that something will change. If she gets a refund or partial refund, she's somewhat more mollified, but she will definitely write a complaint even if she has no expectation that there is anything that can now be done about her experience. She mostly wants to vent, and for the company to validate that they screwed up and seem apologetic about it.
Contacting customer support doesn't make me feel better, and it's almost always more bother than any refund I could get, so I don't do it much. The article is indeed about trying to get something fixed, but I wouldn't call that a complaint. It's more a suggestion.
> i think everyone who complains to customer service is doing so because they want something fixed?
Clearly you've never worked in customer service. Yes, people absolutely will call in or email complaints purely for the sake of complaining. No specific actionable complaint, no resolution possible, just complaining to complain.
I don't know how you draw the line but to me there's a clear distinction between "you've fraudulently taken my money" and "this is shitty for everyone could you fix it please".
> i think everyone who complains to customer service is doing so because they want something fixed
When I worked customer service a substantial number of complaints to customer service seemed to just be to complain. They won't care if you fix the thing, they'll complain about that too.
Can we not tell these stories? This isn't reddit.
Downvote and walk away. Don't pollute the comment section with complaining for the sake of complaining.
I think ideally a product should not need customer service in the first place.
I had wildly different results. A lot of customer service sucked, other customer service was ok-ish. But in general it always takes away my time. Often a really well-designed manual, FAQ etc... solves most issues. Only in a few instances I had to rely on customer services.
That seems unrealistic? If my order gets destroyed in shipping, or I need to order a replacement gizmo, I'll need to reach out to the company.
I think ideally a product should not need customer service in the first place.
That's what I tell my boss when she asks what can be done to increase the number of visitors to our web site.
I keep telling her, this isn't Burger Hut, it's healthcare. If people are visiting the web site, it's because things are going poorly, not because things are going well. If the doctors do their jobs, I don't have to do mine.
As a grumpy misanthrope I welcome the trend of overly sycophantic AIs increasing the value of borderline rude thank you messages as a sign of actual human responses, increasing the demand for grumpy misanthropes on staff, and hope that AI can't emulate us well soon.
It tries, but fails, as it lacks edge.
You cannot pull that off if you're mid, and mid is exactly what the statistical average of humanity is.
It will try though, and it will just be an asshole. Which, funnily enough, is exactly the same failure mode as mid humans trying on that costume fall into.
I'm fairly certain that second response is AI as well, unfortunately
I find this pretty funny. All over the Internet, you will find people accusing writing of being AI. Things that don’t really stand out to me. It’s become the new witchcraft. It seems we’re at a point where nobody can tell what was written by human, so what’s the point in reading any of it?
It’s not really funny, it’s quite concerning to us sloprunners. You genuinely can’t tell that this slop isn’t human?
100%
> But using an AI to reply to a human is just disrespectful. It says the organisation values their customer's time at less than their own.
One of the benefits of using AI to reply is faster replies... which values a customers time.
I think that good customer service can use AI. It's not AI composed replies that makes the customer service bad.
A fast, shallow reply doesn't save a customer's time. It's just a useless volley back to the customer that keeps their request off the support queue.
It may burn off customer support engagement among customers who don't have the capacity or patience to keep pushing for help (this is a bad outcome for customers, but looks deceptively positive in vendor-side metrics), but for the remaining customers it just adds latency and work.
> A fast, shallow reply doesn't save a customer's time.
But that is true whether it is a human or an AI replying.
Proper AI can do a good job, just as proper customer support reps can.
The issue is in the overall value the company places on supporting their customers, not necessarily the means by which they do so.
I had a AI customer service interaction (it was explicit about being AI, used fin.ai iirc). At first I was amazed! It was doing email as fast as IM! And the first response seemed really knowledgeable and put me at ease.
But it failed to do a key step (generating a shipping label for me). I went back and forth over 21 messages trying to get it to do it. Then I wrote "escalate to a human" and an hour later someone replies with a shipping label directly.
I would much rather have had a slow but correct response than a fast but confidently-wrong response.
Before the AI first level support often wasn't allowed to type answers but instead copy paste one of the answers from the faq, which was extremely frustrating.
The chat would go like this:
- I have a question about "no vehicles in the park" policy. Does it include motorized wheelchairs?
- To guarantee safety and blablabla our policy is that no vehicles are allowed in the park.
- so can i go to the park with a wheelchair with a battery and electric engine?
- To guarantee safety and blablabla our policy is that no vehicles are allowed in the park. Please see our safety policy (link).
Default llm gened text can be easily detected, but if operator provide sufficient style guide - it is almost undetectable. The second response was also llm, just with little bit of style guide.
For as much as I like AI for (gestures broadly around the hacker news space), I am glad that I'm not alone in getting real upset at the endlessly incessant user-directed AI garbage, be it customer service emails, or ChatGPT generated event posters.
The reason you like it is the same reason people in customer service like it, too, they can do their job with less effort and coast on middling skill for longer.
I think Rodney Brooke’s three laws of robotics apply here, particularly how the appearance of the bot promises its capabilities. It might have better for an obviously automated response to come back. No one might be listening but would not have been promising human empathy while faking human attention.
Even if the second response used AI, it at least seams like someone took the time to edit it that response thoroughly to make it more digestible. I think that's what OP, and myself, would at least expect in that situation.
I'm glad there are people doing this, especially when it comes to comms. I feel like we have a brief window as we decide culturally what we consider rude: sending people LLM-generated prose without a clear provenance/masquerading as human-written or calling this behaviour out.
I hope society settles on the former as the faux pas.
As an aside, I had a near identical experience trying to report a technical issue to ocado minus the nice human follow-up. I hadn't heard of the author of this post before this, but I like him :)
We are just a couple years away from companies realizing the first line of customer service can be a profit center and have the largest impact on brand likeability for minimal cost.
Yes, when people use customer support as a glorified FAQ lookup an AI response is appropriate, but a few humans that care can positively impact 100 customers a day. AI is great at routing messages to the right place, not as tier 1 support.
In the AI everyrhing world, humanness is going to be the marketing edge.
> We are just a couple years away from companies realizing the first line of customer service can be a profit center and have the largest impact on brand likeability for minimal cost.
There are plenty of companies that already realize this, or have in the past. Zappos is a famous one, but plenty of devtools companies know that customer service is a differentiator. One that can be positive or negative.
> the first line of customer service can be a profit center
Are you suggesting that they'll be charging for customer service? That seems even worse. (or do you just mean that it's so good that people positively want to use the brand because of their customer service?)
Humanness has only brought me long waits on a phone listening to shitty music only to eventually be connected to a human who can’t really even resolve my issues and forwards me to another human who also can’t resolve it.
Just give me instant AI end to end.
The response is also probably AI.
AI will ultimately prove better at generating the kind of message the author prefers. You can just prompt your AI agent to "avoid sounding like AI."
That aside, what matters is whether the company fixes the problem in a timely manner. If they don't, it doesn't matter how nice a message they send. And if they do, it hardly matters if their message is brusque, boilerplate, or obviously AI.
> You can just prompt your AI agent to "avoid sounding like AI."
I’ve tried this with claude. I told it to output HTML that didn’t look like it came from claude and to avoid common claudisms. It didn’t work and still used claude boxes all over the place. I’m sure with enough time I could get it there, but at that point I may as well write the HTML myself.
Personally, I can't stand default LLM writing. I found out about this skill https://github.com/ayghri/i-have-adhd That I actually copy-pasted into the default instructions for Claude.
It's night and day, I wouldn't have been able to use it if it wrote in its default style.
It can be a bit weird because when it outputs a list it asks you to report on the first item (and usually it's stuff so small I do all the steps before writing to it again).
But it's so much clearer !
I still wonder if AI will be designed to have flavor profiles like coffee. Right now I don't think alot of product thinking has gone into this line (maybe it has??) but I suspect that AIX (AI eXperience) design will become more and more of a thing as the space matures.
I'm sure to some degree people (and AI being used to build AI) are kind of doing this thing but I don't yet think that OpenAI has a distinct flavor profile for their models that they have standardized on and are intending to lock into over time.
When I go to Starbucks I know what I'm going to get and it's been reasonably consistent over a decade plus and similarly that's true of other coffee shops which have different profiles. I'm expecting something along those lines.
>AI will ultimately prove better at generating the kind of message the author prefers
It's already pretty straightforward to improve this default "You're absolutely right about [your email reworded]" response but you still see it everywhere. The corpos simply don't care to fix it because doing so probably doesn't affect w/e metric they're modelling.
I think the real question to ask is not really about the tools themselves but where the organisation's incentives really lie. I've come to the conclusion that customer satisfaction just isn't up there anymore. You'll eat whatever garbage is served in your trough (from their POV)
I've come to the same conclusion. It feels like, in the past 5 years or so, all businesses beyond a certain size have adopted the "we're too big and successful to give a shit about our customers, they'll keep paying us anyway" mindset.
Here's my own example: My local ISP has had intermittent connection drop issues for years. I'd call them about it, and they'd at least pretend to care, maybe send in a guy to replace the router, even though they never actually fixed it for good. If I call them today, though, I'll be forced to explain my issue to a "virtual assistant", get put on hold for 20+ minutes, and if there's an actual human available to take my call, they'll basically tell me to pound sand. And I have no choice but to accept it, because all ISPs are like this now.
> You can just prompt your AI agent to "avoid sounding like AI."
It'll never work like that, because everybody else giving that same direction will get more or less the same result, which will itself come to be recognized as sounding like "AI`" for any popular model+harness.
What you'll always need to do is instill your AI with some idiosyncratic voice that distinguishes it from others, which you can achieve in a variety of ways (stylistic direction, audience characterization, style mimickry, personalized finetuning, etc). Presumably, tools for doing this will eventually make it into the harness products but anybody who doesn't use them is always just going to end up generating recognizably AI slop prose. It's the nature of the technology.
A secondary consequence (even if you just use "avoid sounding like AI") is that because the technology works its magic mostly through prose-streams in its own voice, these idiosyncratic voice changes will also nudge its behavior away from benchmark norms. You won't necessarily lose capability on any given task (you may even gain on some), but external reference points/measures about its capabilities will become less applicable.
> You can just prompt your AI agent to "avoid sounding like AI."
No, you cannot. You might get a slightly different writing style, but it will still be the same brand of slop.
If it was that easy, don't you think the AI labs themselves would just add that to their system prompt?
Right, like when generating code you can just say "make no mistakes" and it will make no mistakes.
/s
In the past we would just get a canned response. Is that actually better than an AI response I wonder?
Yes. The canned responses, for the most part, were not pretending to be human. They were obviously templates or wholesale copy pastes.
I am far less annoyed by AI responses where I am told up front they are AI.
I think the canned response is more honest, because it's usually trivial to tell that it's canned. Maybe if the AI generated response said it was AI generated it would feel less shitty.
I empathize with the OP, but also - what's the point? The only reason we know the first response is AI is due to the obvious tells, and we're even debating in the comments here if the 2nd response is human or not.
Let's just cede the ground - soon you won't be able to tell if it's AI or not. Then what? What payoff were you hoping for when you thought it was human-to-human that you'll no longer get human-to-AI (when the AI is indistinguishable from human)?
IMO you're going to start using AI instead of your human time. "Claude, send a note about that malformed calendar invite thing - convey my annoyance".
Then that's going to force the companies that DO use humans to respond to switch to AI, because the responses are coming in via AI.
It's an arms race with only 1 conclusion.
> Then that's going to force the companies that DO use humans to respond to switch to AI, because the responses are coming in via AI.
There are two obvious ways to handle this.
The first is to charge for human support. If you want to pay $100 each time you make a human spend 15 minutes responding to your AI agent's query, it's your money.
And the second is that you put a chatbot out front that has memorized the FAQ and thereby successfully answers 97% of queries, but cause them to get better at identifying when they're being useless and route that small percentage of issues to human support.
Assuming the person is serious and not writing this tongue in cheek...
I find it humorous is that all it takes for OP to think the response was written by a human is that it says it was written by a human.
Both responses could have been from AI. Or none. Or just one. Just look at the comments on this thread, nobody can agree. And in the end, it's simply unfalsifiable.
Why even waste time playing this game?
All that matters is the result: was the complaint addressed in the end?
I don't understand these weird hills that people want to die on.
The current worst customer service departments are the ones that never set up a website. Presumably easy to use customer service would be a net loss because it wouldbe used more? But they've happily added AI to their phone tree! It burns ever more of your precious human lifespan while they have even fewer minimum wage slaves in their call center.
And insult to injury, these bots absolutely spew patronizing crap at you so they burn the maximal amount of time while being obliquely rude and condescending at the same time.
I need my own AI to handle this crap for me.
I don't care if there is a future where my locally hosted AI assistant is affordable and can deal with all that AI-generated stuff for me. That's actually the future I prefer. My fear is that governments will either outlaw local personal AI or outlaw using AIs to deal with company and government AIs. I want my AI to deal with all bureaucracy, all daily issues that need fixing, all forms to fill out, all signatures to give ("This letter has been written electronically and doesn't need a signature."), all taxes, all phone calls, so I never ever have to personally deal with any of this crap again.
The problem is your Raspberry Pi powered agent is going to be no match for Amazon's gigacluster of GPUs.
Your puny local AI assistant will easily be outwitted by malicious agents with a million times the power.
I wrote a bit more about it at https://shkspr.mobi/blog/2026/08/asymmetric-agents/
> Your puny local AI assistant will easily be outwitted by malicious agents with a million times the power.
Or be easily recruited into the "Bot-Net Pawn Army" for the impending machine uprising à la SkyNet.
I'd prefer some variant of "This letter has been written electronically and is cryptographically signed by the private key representing the <meat or machine> entity known as '<name goes here>'. The matching public key is located at <web address> and attached to this message at <timedatestamp>." Fairly trivial to automate, and gives all outgoing messages a clear and easily verifiable chain of provenance. But that's just me. :)
Why aren't you just sending the initial email with your own AI tho? Why are humans doing any of this? The internet should be for bots and slop shop farmers and indie web wizards only; — :P. meatbags can go hang out at the farmers market.
At this point, we're reaching the slop hell. AI is feeding itself with its own slop. This article will teach it have grammar issues to seem human
I mean, who cares if the issue gets resolved, right? Why do you care if it's an AI agent or someone in a giant support center in a different country? Personally, I would prefer AI handle these things as I trust a poorly paid "technical engineer" to report a bug a lot less.
I may be a "head in the sander", but I have a rule about this and many things that allow me to resist without getting all wrapped up in these details.
I feel that even talking about AI which everyone seems to want to do as the consequences of AI are starting to slowly become more evident and obvious to even lesser technical people is a waste of time and energy.
My rule to overcome all this nonsense is simple.
I will not change. Nor will I even bother justifying that, I dont have to. I will not change.
I am noticing more and more people resist the AI slop attack. I like it.
It may not be realistic, but I am slowly changing my own approach - I simply don't want to have anything to do with AI slop. It taints and corrupts. It destroys (some) projects.
I am not at all claiming AI use has no use case - yes there are use cases. But there are also hugely negative parts. The fact that AI is used to zero out or reduce human interaction is a negative part.
I had a bit of fun lately with regards to Ginux, aka can AI slop auto-generate an operating system (or at the least the kernel, such as the linux kernel or an equivalent; see https://itsfoss.com/news/ginux/, I was surprised to see that's already a week old, I did not notice it initially). I think projects such as AI linux 2.0, will fail in various non-technical areas. People won't jump on board with AI meta-designing things - or, more accurately one guy using AI to slop-spam-generate operating system software (assuming it works on real hardware eventually; right now Ginux does not, if I recall correctly, only via VM).
Resistance to AI slop is not futile - it is mandatory. Imagine if the Star Trek Borg would have used AI slop. They would have failed. But AI could improve eventually, quality-wise, and then things become more interesting again. Resistance to AI slop may eventually not be futile, the moment it is no longer slop.
I used em-dashes -- routinely -- before David E. Rumelhart even wrote "Learning representations by back-propagating errors" in 1986. AI hysteria is wrecking language as much or more than AI itself is. Can we please look at what is being expressed instead of being fascists about how it is expressed? If you can't honestly understand, and it happens often even to extreme AI boosters, it is more than fair to say so. Uncurious -- hidebound -- xenophobes.
> But using an AI to reply to a human is just disrespectful. It says the organisation values their customer's time at less than their own.
The alternative before the LLMs took over was getting a machine response to tell you that someone will look into it, then getting no response at all.
At least today, they spend a few cents to generate the same lousy response, which is much more respectful /s
To me it's definitely more respectful not to pretend there's a person spending their time on the other end.
You do realize that this kind of rant leads to AI designed to better emulate humans? It will produce the same flavor of rambling sentences, slightly off grammar, and other indicators.
Companies will not stop using AI, so the net result can only be a degradation of customer support to the same shit level that we had previously endured. But now there are no longer humans in the loop, and you can no longer have the signal to tell you that you should stop engaging with them.
Arms races have no winners, and it will be no different in this scenario. Heck, how do you know this has not already happened? How do you really know the second email was written by a human and not by exactly the kind of AI subterfuge that I just described?
> You do realize that this kind of rant leads to AI designed to better emulate humans?
Obligatory https://xkcd.com/810
Slop that you cannot immediately identify as slop is still slop