> Anthropic is an AI company that builds one of the most capable AI assistants in the world. Their support system is a Fin AI chatbot that can’t actually help you.
This really cuts to the reality of AI hype: no, agents are not nearly as capable as OpenAI, Anthropic, etc. need you (or rather your C-suite, itching to fire you) to believe. They really, really need you to believe the hype. How can you tell? Cases like this and the fact that there are 5000 open bugs, constant regressions, ignored feature requests in the CC repo. The fact that Codex doesn't fully implement the simple and well-defined MCP spec for prompts. The fact that even CC has gaps with the MCP implementation...a spec that they created!
If the progenitors with functionally infinite tokens can't get this basic stuff right, everything else they are doing is just blowing smoke. I don't care if you can ship a kernel compiler or a janky browser; how about just make your software work? The smartest guys in this space, engineers making 7 figures in TC, with billions in capital, unlimited tokens, and access to the best models cannot make a simple customer support chatbot work.
But you! You're expected to deliver that customer support agent that's going to allow them to cut 500 people from payroll. You'll have it by Monday, right?
What if they built their company with poor support, so they don’t have to hold up to any standard ? But others companies have historically good reputation for good customer support, and maybe AI can help them automate easily 80% of easiest requests
It's really a bit fascinating. I've had Claude one-shot complex functionality... and I've had it be unable to debug its own .mcp.json file effectively.
In all seriousness, shouldn't Anthropic be heavily dogfooding this sort of use case? I'm also not a huge fan of Amazon's support system, but they at least seem to be using their AI tools a lot for support responses (which has its own issues, but credit where its due).
Every conference talk on this stuff seems to suggest that we're all way behind the curve on AI implementation, but I suspect its mostly smoke and mirrors and mechanical turks. My company invests heavily in automated IVR and chat responses and we still optimize for getting the customer to a real agent. Those agents are largely overseas BPOs, but at least that's better than an AI loop that gets you nowhere.
I've also been waiting over three weeks to speak with customer support after being gifted an annual subscription just as my payment card expired. The failed payment (after the $200 gift) downgraded my account to the free tier and I lost my annual subscription. I had to pay another $20 to get back into the pro tier plan, but now for some reason I only have $197 in credits and I'm on the monthly subscription instead of the annual. Anthropic basically just made 3+ months of credits disappear for their own billing mistake.
The kicker? When you get downgraded to the Free tier, they don't offer any support beyond the AI bot. You have to go through some hoops to get it to open a support ticket to maybe talk to a human in 4-5 weeks. Unbelievable.
I had the displeasure of interacting with that support agent earlier today and was very surprised. It's just as good as the one my ISP has.
We're meant to trust Anthropic enough to replace all of our engineers by their model for writing our software but somehow they don't trust it enough to let it handle simple customer support decisions. But shhhh, it's voluntarily nerfed just slightly bellow ASI for our safety.
OpenAI, xAI, Anthropic, Google, MSFT, Spotify, Duolingo and NVidia - those are the ones that come immediately to mind. They're either selling the AI (or the tools to make the AI) or hoping against all hope that they're on the right side of bubble history.
If we soften the claim to "increase engineer productivity" I think something like 70% of engineers would also agree. If you tack on "if applied wisely" then you'll probably be up to 95% of engineers
It's important to remember that a chargeback should be considered the nuclear option, and, when using it, one should be comfortable with the possibility that one might never do business with this company again, since it could result in being blacklisted (even if one is, in fact, in the right). I'm not saying not to do it, but one should keep in mind the potential repercussions.
If a business attempts to steal from me I instantly charge back and the onus is on them to prove that I owe them money. I do this all the time and have never been blacklisted.
I don't think it's helpful to think about this as the company "trying to steal from you". There is no intention here. It's just something that got lost in a bad IT system. You gain nothing from issuing a chargeback. You imperceptibly nudge some statistic and a "banned for life" flag might automatically get flipped in a database. There's no righteous comeuppance here.
You try to contact support, pester them a bit, call someone if possible, and eventually, you may get your money back. If you don't, then you issue the chargeback.
What exactly is being blacklisted in this case? They have a credit card and sort of an address/email. These aren’t exactly hard to “regenerate”.
The risk with something like a google account is more the amount of shit people have tied up in the account, that doesn’t sound like it would be the case here.
More generally in this age it seems like disputes (which don’t necessarily result in a true chargeback) are just more routine than nuclear in the US. I now do them routinely for certain restaurants with delivery and a customer service contact that is essentially a black hole (it’s usually a partial dispute). They’ve obviously figured whatever CB rate they’re getting is cheaper than CS.
I’ve never been “blacklisted” in 20 years including Amazon.
waiting for month for a refund (and having lost access to the pro plan immediately but no immediate refund) is definite grounds for chargeback.
there is no human on the other end of the chain, and I bet that chargebacks are how they issue refunds (ie relying on the "nuclear" option as the standard practice of how refunds fundamentally works at their company.
ie "don't need to answer emails about refunds, because if they really wanted their money back, they'd issue a chargeback" as part of the regular procedure.
a lot of companies do this, and it's a common way of minimizing customer support budgets.
I always wondered about this. Do companies tie the credit card to an identity to block or do they just block the cc number?
If the latter, seems like a small friction point for a consumer. Given how often cc numbers change and how many an (American) consumer has, this won’t block anything unless you are charging back more than once every few months.
More like, you don't sue a vendor and then expect the relationship to go back to status quo ante.
A chargeback is essentially binding arbitration and it can be existentially costly for small businesses, especially those unable effectively to advocate for themselves in a fairly complex and little-known process. Excess chargeback initiations - even of failed chargebacks - will also get acquirer accounts closed, meaning the business formerly a client of that acquirer can now no longer accept credit cards. (Modern acquirers like Stripe also do this, because the card issuers and payment networks will eventually cut them off if they don't: Stripe is not "too big to fail" according to Visa, which is why you may not sell sex or porn via Stripe.)
Anthropic doesn't need to care, of course. No one is going to fire them as a customer over excess chargebacks, and a hundred such fees are still cheaper than one hire. Anthropic has a burn rate. Chargebacks impinge much more heavily on businesses that need to earn money selling goods or services. It's important not to confuse one with the other.
TBF, I think Anthropic is a victim of their own success right now. We've had clients reach out to their sales team and be unable to reach anyone. I think they are just busier than they can actually handle.
Yes, it’s pretty much the case, they are trying to scale as fast as they can from what I understand. Their growth over the last year has been just insane
Have you tried suieng then in small claims court? They skimp in being a real company with real legal support by burning infestor capital, because staff attorney salaries are accounted for much harder than individualized lawsuits from practices not directly resolved next lay period.
Most people who commit wire fraud weren't socially bullied and criticized enough before their professional positions to keep in line legally. Useless failures.
Anthropic doesn't allow you to hide or unshare Projects which were shared by team members who are no longer on the team. Contacted them about this two months ago, have yet to hear from any human.
Sorry to hear that. Yeah, it seems like this is a shared experience among many Claude users. Hoping that this post will draw more attention to the issue so that Anthropic will address it.
This is the risk of being a consumer in the AI world - companies are running extremely lean on real humans and are deferring support to AI chatbots with no real reasoning abilities...
Also an issue with scale - for example, Google having similar issues of not handling small, isolated cases.
See the other commenter/thread that recommended I do this. I'm worried that by doing a chargeback, I will be blacklisted from using Anthropic's services, which I feel like is a reasonable assumption.
Until they do start doing identity verification, I think you're good. Frankly, don't be a coward. If you're getting treated like this, why would you even want to use their services in the future?
That will get their attention - to blacklist you from ever doing business again with them. People saying this is a nuclear option are telling this because they know what a charge back means for a business owner. So treat it like that.
I guess I shouldn’t feel so bad then that I have a ticket open that I keep updating every few days with how long it’s been without a response. It’s only been a few weeks.
I'm not surprised, I burn (on purpose) more than 15k$/month on Anthropic tokens and I've never been able to talk to any of their sales despite filling the contact form every week for the past 4 months :')
Well, that's kinda the problem, isn't it? Even after being erroneously charged and ghosted by their non-existent support for a month, you'll still happily keep paying for their services.
If most people think like you, why indeed bother providing support at all?
Good point. I did actually cancel my Claude subscription a week or two ago, but I renewed it (regretfully) just the other day. The only other SOTA model that seems to be on-par with Opus 4.6 for engineering work is (maybe?) Codex 5.3, though I would rather not support Sam Altman indirectly.
I wouldn't hold your breath. It seems like the only way to get an actual human response is by complaining on Twitter/X and hoping that Boris Cherny responds. https://x.com/bcherny
> I even wrote it not only per email but also in the "in-chat feedback" system
Yeah, I did the same. Before falling back to sending an email to support@mail.anthropic.com (which my blog post references), I had 3 separate Fin AI in-chat convos trying to get in touch with someone. All of them defaulted to the "ask for a refund" workflow that only applies for subscriptions and left me more frustrated than anything.
I did a chargeback against OpenAI for something similar and I showed my credit card company the logs with the support bot, as it was my only point of contact for the company.
Thinking it might be time to push for some laws to mandate companies have better systems to handle and address concerns that impact customers businesses and livelihoods.
This inability to reach and/or get things resolved through customer support channels seems endemic, and probably generally part of the enshittification trend as a whole.
Ah, I wouldn't have written this blog post if I had known that that was the usual turnaround time. There should really be more transparency on when one should expect to hear back rather than the generic response of "a member of our team will be with you as soon as we can."
edit: albeit another commenter claims they have been waiting for 2 months...
Large corporations have been downsizing on QA and CS roles since before the LLM era. For many of those companies the lack of proper QA leads to more problems for users which compounds the lack of available CS staff. It's called either enshittification or maximizing shareholder value, can't remember which.
Once again [0], Anthropic does not care about you and they are not your friends.
The other day Dario and Co, were looking at a robotic lamp that does your laundry and folds your clothes. He cares more about investing in that than your billing issue.
To them, they see us as gambling addicts, whilst we pay them their overpriced credits at their casino.
This sucks but is not surprising at all. Anthropic has more demand than it could ever fulfill, and looking into support tickets asking for refunds is never going to get anyone’s attention. If you actually want the money back, assuming you live in the US, this is what small-claims court is for.
If the progenitors with functionally infinite tokens can't get this basic stuff right, everything else they are doing is just blowing smoke. I don't care if you can ship a kernel compiler or a janky browser; how about just make your software work? The smartest guys in this space, engineers making 7 figures in TC, with billions in capital, unlimited tokens, and access to the best models cannot make a simple customer support chatbot work.
But you! You're expected to deliver that customer support agent that's going to allow them to cut 500 people from payroll. You'll have it by Monday, right?
Let that sink in.
What if they built their company with poor support, so they don’t have to hold up to any standard ? But others companies have historically good reputation for good customer support, and maybe AI can help them automate easily 80% of easiest requests
It's really a bit fascinating. I've had Claude one-shot complex functionality... and I've had it be unable to debug its own .mcp.json file effectively.
In all seriousness, shouldn't Anthropic be heavily dogfooding this sort of use case? I'm also not a huge fan of Amazon's support system, but they at least seem to be using their AI tools a lot for support responses (which has its own issues, but credit where its due).
Every conference talk on this stuff seems to suggest that we're all way behind the curve on AI implementation, but I suspect its mostly smoke and mirrors and mechanical turks. My company invests heavily in automated IVR and chat responses and we still optimize for getting the customer to a real agent. Those agents are largely overseas BPOs, but at least that's better than an AI loop that gets you nowhere.
I've also been waiting over three weeks to speak with customer support after being gifted an annual subscription just as my payment card expired. The failed payment (after the $200 gift) downgraded my account to the free tier and I lost my annual subscription. I had to pay another $20 to get back into the pro tier plan, but now for some reason I only have $197 in credits and I'm on the monthly subscription instead of the annual. Anthropic basically just made 3+ months of credits disappear for their own billing mistake.
The kicker? When you get downgraded to the Free tier, they don't offer any support beyond the AI bot. You have to go through some hoops to get it to open a support ticket to maybe talk to a human in 4-5 weeks. Unbelievable.
I had the displeasure of interacting with that support agent earlier today and was very surprised. It's just as good as the one my ISP has.
We're meant to trust Anthropic enough to replace all of our engineers by their model for writing our software but somehow they don't trust it enough to let it handle simple customer support decisions. But shhhh, it's voluntarily nerfed just slightly bellow ASI for our safety.
You're not meant to trust. Stop getting hooked on company PIR
They didn't ssy they did trust their claims.
Who keeps claiming these models are meant to replace engineers?
OpenAI, xAI, Anthropic, Google, MSFT, Spotify, Duolingo and NVidia - those are the ones that come immediately to mind. They're either selling the AI (or the tools to make the AI) or hoping against all hope that they're on the right side of bubble history.
If we soften the claim to "increase engineer productivity" I think something like 70% of engineers would also agree. If you tack on "if applied wisely" then you'll probably be up to 95% of engineers
the remaining population of linkedin users?
I tried their Pro plan on March 1 and immediately noticed how bad their usage limits were, so I asked for a refund that same evening.
Their chatbot accepted the request, I was downgraded to the free plan immediately, and since then I have been waiting for the money.
Did you follow up? You might need to do it again before charge back.
Thankfully that's not Google, so your life is not going to be turned upside down because they don't give a f*.
I opened a new ticket over three weeks ago to ask about the status of the refund, and that has been left untouched as well.
Now I have submitted a reclamation request to my bank and am waiting for a response.
Yikes. That's unacceptable. Crazy that it has been over a month and you still haven't gotten the refund.
Issue a chargeback.
It's important to remember that a chargeback should be considered the nuclear option, and, when using it, one should be comfortable with the possibility that one might never do business with this company again, since it could result in being blacklisted (even if one is, in fact, in the right). I'm not saying not to do it, but one should keep in mind the potential repercussions.
[delayed]
If a business attempts to steal from me I instantly charge back and the onus is on them to prove that I owe them money. I do this all the time and have never been blacklisted.
I don't think it's helpful to think about this as the company "trying to steal from you". There is no intention here. It's just something that got lost in a bad IT system. You gain nothing from issuing a chargeback. You imperceptibly nudge some statistic and a "banned for life" flag might automatically get flipped in a database. There's no righteous comeuppance here.
You try to contact support, pester them a bit, call someone if possible, and eventually, you may get your money back. If you don't, then you issue the chargeback.
Yeah that kind of seems like antiquated fear-mongering. Next they should call the BBB and leave a strongly-worded review!
wait, int the BBB just boomer yelp?
What exactly is being blacklisted in this case? They have a credit card and sort of an address/email. These aren’t exactly hard to “regenerate”.
The risk with something like a google account is more the amount of shit people have tied up in the account, that doesn’t sound like it would be the case here.
More generally in this age it seems like disputes (which don’t necessarily result in a true chargeback) are just more routine than nuclear in the US. I now do them routinely for certain restaurants with delivery and a customer service contact that is essentially a black hole (it’s usually a partial dispute). They’ve obviously figured whatever CB rate they’re getting is cheaper than CS.
I’ve never been “blacklisted” in 20 years including Amazon.
waiting for month for a refund (and having lost access to the pro plan immediately but no immediate refund) is definite grounds for chargeback.
there is no human on the other end of the chain, and I bet that chargebacks are how they issue refunds (ie relying on the "nuclear" option as the standard practice of how refunds fundamentally works at their company.
ie "don't need to answer emails about refunds, because if they really wanted their money back, they'd issue a chargeback" as part of the regular procedure.
a lot of companies do this, and it's a common way of minimizing customer support budgets.
Unless you're big cheese, too many disputes can get a company cut off. Disputes aren't free to mediate, there's a cost to handle each one.
Visa/MC can block a company, happens for lots of reasons.
I always wondered about this. Do companies tie the credit card to an identity to block or do they just block the cc number?
If the latter, seems like a small friction point for a consumer. Given how often cc numbers change and how many an (American) consumer has, this won’t block anything unless you are charging back more than once every few months.
So the Anthropic company would blacklist you for taking your money back by force that they owe you?
Ok sounds like evil should be labeled and not tolerated as anything else.
More like, you don't sue a vendor and then expect the relationship to go back to status quo ante.
A chargeback is essentially binding arbitration and it can be existentially costly for small businesses, especially those unable effectively to advocate for themselves in a fairly complex and little-known process. Excess chargeback initiations - even of failed chargebacks - will also get acquirer accounts closed, meaning the business formerly a client of that acquirer can now no longer accept credit cards. (Modern acquirers like Stripe also do this, because the card issuers and payment networks will eventually cut them off if they don't: Stripe is not "too big to fail" according to Visa, which is why you may not sell sex or porn via Stripe.)
Anthropic doesn't need to care, of course. No one is going to fire them as a customer over excess chargebacks, and a hundred such fees are still cheaper than one hire. Anthropic has a burn rate. Chargebacks impinge much more heavily on businesses that need to earn money selling goods or services. It's important not to confuse one with the other.
TBF, I think Anthropic is a victim of their own success right now. We've had clients reach out to their sales team and be unable to reach anyone. I think they are just busier than they can actually handle.
Yes, it’s pretty much the case, they are trying to scale as fast as they can from what I understand. Their growth over the last year has been just insane
Have you tried suieng then in small claims court? They skimp in being a real company with real legal support by burning infestor capital, because staff attorney salaries are accounted for much harder than individualized lawsuits from practices not directly resolved next lay period.
Most people who commit wire fraud weren't socially bullied and criticized enough before their professional positions to keep in line legally. Useless failures.
Anthropic doesn't allow you to hide or unshare Projects which were shared by team members who are no longer on the team. Contacted them about this two months ago, have yet to hear from any human.
Sorry to hear that. Yeah, it seems like this is a shared experience among many Claude users. Hoping that this post will draw more attention to the issue so that Anthropic will address it.
This is the risk of being a consumer in the AI world - companies are running extremely lean on real humans and are deferring support to AI chatbots with no real reasoning abilities...
Also an issue with scale - for example, Google having similar issues of not handling small, isolated cases.
Hope you get your money back!
Google is like this since ever, way before AI, so no, that's not the reason.
You can call Google and their support for business customers is personal and excellent.
Thanks, I hope so too!
If it's a money issue then file a charge back with your credit card. That generally gets someone's attention.
See the other commenter/thread that recommended I do this. I'm worried that by doing a chargeback, I will be blacklisted from using Anthropic's services, which I feel like is a reasonable assumption.
Until they do start doing identity verification, I think you're good. Frankly, don't be a coward. If you're getting treated like this, why would you even want to use their services in the future?
> why would you even want to use their services in the future?
Uhhh my base case is you will be forced to or just be forgotten, not unlike not having a cell phone or a bank account.
Did you get the API credit? Maybe it's a wash?
That will get their attention - to blacklist you from ever doing business again with them. People saying this is a nuclear option are telling this because they know what a charge back means for a business owner. So treat it like that.
> I also wanted to confirm with a human on exactly what went wrong
They wouldn't be able to tell you. The entire back end system is probaby vibe-coded and nobody really understands what it does.
I guess I shouldn’t feel so bad then that I have a ticket open that I keep updating every few days with how long it’s been without a response. It’s only been a few weeks.
I had a similar thing happen where I was looking to recover funds from unexpected extra usage charges and got went through an identical experience.
I realize the company barely has time to cash checks, but failing to handle small fry reasonable charge disputes should be handled appropriately.
I don’t know why you waited so long to submit this to the support forum they actually read, which is of course this one.
You people sound like Token Addicts. No matter how bad the Ai treats you, you keep coming back. Why should they ever change?
TBF I'd probably pay some solicitor $50 to have them send a nicely worded letter after 2 weeks.
You're too kind for the company trying to steal from you - whether intentionally or by negligence, doesn't really matter.
Or the small claims court mentioned by someone else. Make sure to add your time and the cost of the representation.
I'm not surprised, I burn (on purpose) more than 15k$/month on Anthropic tokens and I've never been able to talk to any of their sales despite filling the contact form every week for the past 4 months :')
You're worth a whole department of claude subscribers which tells me they don't give a fuck about API users.
$15K/month? USD? Are we talking company funds or personal?
This is what credit card chargebacks are for.
I'm sure this guy would like to actually keep using Claude though instead of getting permanently banned.
Yep. I don't want to get blacklisted from using Claude indefinitely by doing a credit card chargeback.
What happened to voting with your wallet.
Well, that's kinda the problem, isn't it? Even after being erroneously charged and ghosted by their non-existent support for a month, you'll still happily keep paying for their services.
If most people think like you, why indeed bother providing support at all?
Good point. I did actually cancel my Claude subscription a week or two ago, but I renewed it (regretfully) just the other day. The only other SOTA model that seems to be on-par with Opus 4.6 for engineering work is (maybe?) Codex 5.3, though I would rather not support Sam Altman indirectly.
Then get fucked in the wallet I guess ?
Use another CC and email address?
Stripe's pretty good at using other signals to block this sort of thing.
I didnt know that they have any useful support at all! :-D
I sent them some feedbacks one some issues, actually good ideas, and I didnt get any response so far.
I wouldn't hold your breath. It seems like the only way to get an actual human response is by complaining on Twitter/X and hoping that Boris Cherny responds. https://x.com/bcherny
I even wrote it not only per email but also in the "in-chat feedback" system (you can add a text to a response)
Also on LinkedIn they are siltent - I reached out to one of their sales reps, no response.
Maybe in the end we will have "Google-class" support?
> I even wrote it not only per email but also in the "in-chat feedback" system
Yeah, I did the same. Before falling back to sending an email to support@mail.anthropic.com (which my blog post references), I had 3 separate Fin AI in-chat convos trying to get in touch with someone. All of them defaulted to the "ask for a refund" workflow that only applies for subscriptions and left me more frustrated than anything.
I did a chargeback against OpenAI for something similar and I showed my credit card company the logs with the support bot, as it was my only point of contact for the company.
Thinking it might be time to push for some laws to mandate companies have better systems to handle and address concerns that impact customers businesses and livelihoods.
This inability to reach and/or get things resolved through customer support channels seems endemic, and probably generally part of the enshittification trend as a whole.
Their response time is usually around a month IME, yes.
Ah, I wouldn't have written this blog post if I had known that that was the usual turnaround time. There should really be more transparency on when one should expect to hear back rather than the generic response of "a member of our team will be with you as soon as we can."
edit: albeit another commenter claims they have been waiting for 2 months...
Just enough time to have your chargeback denied by the bank.
Large corporations have been downsizing on QA and CS roles since before the LLM era. For many of those companies the lack of proper QA leads to more problems for users which compounds the lack of available CS staff. It's called either enshittification or maximizing shareholder value, can't remember which.
Why not both? ;)
Fin is actually Intercom’s branded agent, so if Anthropic is using their own model for support at all isn’t clear.
I was curious so I looked into it. Seems like my issues were encountered when Fin AI was running on Sonnet 4.0, though Fin AI's new model (Fin Apex 1.0) was rolled out ~2 weeks ago. https://www.intercom.com/blog/announcing-fin-apex-the-age-of...
if this is on a credit card you can get the money back from the credit card company for "undelivered goods"
Once again [0], Anthropic does not care about you and they are not your friends.
The other day Dario and Co, were looking at a robotic lamp that does your laundry and folds your clothes. He cares more about investing in that than your billing issue.
To them, they see us as gambling addicts, whilst we pay them their overpriced credits at their casino.
The house (Anthropic) always wins.
[0] https://news.ycombinator.com/item?id=47679322
"Okay, Claude, and this time we're going to slice the salami really thin, because this sandwich is for all the marbles..."
This sucks but is not surprising at all. Anthropic has more demand than it could ever fulfill, and looking into support tickets asking for refunds is never going to get anyone’s attention. If you actually want the money back, assuming you live in the US, this is what small-claims court is for.
The fact that denizens of HN think that taking a company to small-claims court is a reasonable approach to getting refunds :: SMH