Ask HN: Will low quality AI customer support be the new normal?
Now whenever I reach for customer's support chat or phone I get an AI agent replying to me and I get into a useless loop for a couple of minutes before I start begging it to link me to a real person.
Will companies start losing customers because of that or people will eventually get used to this?
Yes. Because, the incentives for companies are too big to ignore. The math being done right now is $10/hr for voice support in (say) the Phillipines is much more than $0.10/hr having an AI do it, even factoring in the cost of some customer churn. And the risk of the latter for some services can be discounted to zero if the user has no viable alternative.
Second, the best support line is one where no one calls in. So there are strong incentives to make it hard to do so, either positive (create good self service options) or bad (hide the number, hide emails, etc).
If you don't want this future then make sure you always ask for the human and constructively provide feedback that you will penalize any supplier that relies on it.
We are already in a K shaped economy wrt call centers. Being a very frequent Hilton hotel guest, say, gets you a separate call center number, US based live human support, people who can solve your problem, etc. Not being a Hilton member gets you the 45 minute wait and a bot.
There's always a viable alternative. But not for customers who are always looking for the cheapest viable products or services.
So those customers will spend too much time and frustration with bad customer service, because they think it's outrageous to consider a competitor who is a little bit more expensive.
At $PREVIOUS_EMPLOYER, they prided themselves on having the best support operators, with deep technical prowess of their specific knowledge domain. Their L1 were better than some other business' L3s infrastructure specialists. This lasted for more than a decade, right until a merger with another business that also had their own human support department and certain "strategic opportunities" were identified. At first, some operations were consolidated, and some people were fired. Once the corporate machine saw the numbers, they just kept going. Nowadays, its almost full AI.
The real fun one is calling the utility company to report a downed line from a pole snapped in half and the bot agent doesn’t transfer you to a human even when you say “emergency”, “public hazard”, “utility pole obstructing a roadway”, “potential property damage your corporation will be liable for”, “immenient danger to life and limb” or any and all applicable keyword variations that might warrant humanly answering (and, to add insult to injury, cannot take a report of the event either but simply ends 10 of your calls in a row with a texted link to livechat with its textbot sibling). Like +−×÷ says nearby, it’s austerity.
I believe it will, indeed, become the new standard as it’s too cost efficient for the companies but they may loose a lot of people that are not tech savvy on the way (e.g. older people). This will either drive their customers to competitors properly managing support or force the companies to revisit the way they handle support.
I think eventually you probably will not able to tell whether support is ai or human. models and agents will only continue to improve, and customer support is one of the clearest business use case.
For companies, ai is cheaper, scalable and does not require the same onboarding process and training process. so in that sense, yes, it probably becomes the new normal.
My guess is in the long term, we will have something like a hybrid model. Where ai answers repetitive questions and straightforward workflows, while humans can answers the nuance.
I think it will be, but only for sometime. As the AI improves, the process also will improve and adapt overtime. And hopefully support will be much more smarter and easier and not low quality..
I asked Amazon for help with my account. They couldn't figure out the URL to get me where I needed to go. They said they needed to escalate. The question was definitely not complicated. They wanted to call me asking questions.
Meanwhile, Gemini gave me the url and explained what I needed to know in one reply. Problem solved. Same question to ChatGPT gave me the correct answer as well. I bet Claude or Grok would have also found the correct answer.
This is where things take a turn: Google.com/ddg/bing failed. amazon's own search failed. Its worth knowing that the good results from the chatbots pointed at Amazon's web.
Based on this one example, I could begin to think document style search engines are dead.
Presumably Amazon were trying to put me into their little Voice AI microcosm. They'd obviously (?) record my voice (yes) of me getting increasingly annoyed (likely) at some Voice Chat Bot (likely) and then they'd apologize (?) for not knowing (probably?).
Is that the goal? Waste 30 minutes of my time for what could have been answered by a chat bot but not a web search?
The underlying situation seems even weirder than simply trying to cut costs. I think Amazon has cut so many corners they have forgotten what shape they are trying to create.
I have only one answer so far: Their customer support has no idea how to navigate their own website and this is by design. This is NOT the customer support people at fault. Let that all sink in: Someone got a bonus for this service design. This is somehow optimal according to Amazon. Yet this style of service design will only get worse for paying customers. It will be much worse for non-customers.
Progress!
I don't think this is restricted to Amazon either. I think this is industry wide. The in-page chat bots are usually just as broken.
AI is essentially austerity. We're going to get synthetic/inferior versions of many products and services, wherever using AI can make them cheaper to produce. Companies no longer fear losing customers because customers have proven so passive and docile as these forces have already started to accelerate over the last decade.
AI customer support is spotty in places but is improving rapidly every month. Humans aren't getting better every month. Also it's 100x cheaper than hiring a human, and has a bunch of other built-in advantages, including
1. Native understanding of most languages
2. Full knowledge of everything in the company's knowledge base
Voice agents are a royal pita. They have trouble with any kind of British regional accent, they are tiny models because anything usable is too slow, and they have zero capability because they'd probably mangle the toolcalls anyway.
Elevenlabs using a completely incapable voice agent for sales outreach made me wonder if they were actively trying to dissuade customers.
It feels like the sort of thing a disinterested engineer implemented to appease a manager fixated on the shiny.
Right now cheap AI customer support is genuinely bad and does hurt the customer experience. But this looks like a transitory phase, same trajectory as text LLMs, which got dramatically better while getting cheaper over a couple of years.
Give it 12-18 months and for routine support flows the gap between a well built voice agent and a human will be hard to notice. Complex or emotional cases are where humans stay, which is fine, that’s where they add value anyway.
We better be prepared for armies of agents in companies... not only for customer support but everywhere. For now it may seem that they are not quite reliable, but with the right context and memories and the right number of humans behind to manage them (unmanaged agentic AI seems too optimistic for me), the quality will increase dramatically and we'll no longer have humans on the loop. Even ourselves we'll probably let our personal agents solve our own problems, so it'll be agent-to-agent calls and interaction.
Deflection of low priority and low human resource requirement will continue to need such approaches. Those happen even without AI now but the way. It’s just too costly to have humans pickup a call for something that was knowledge related (deflect to knowledge and FAQ) or self service (what’s my outstanding balance) at scale
I believe customer support should actually be done by humans, as customers feel undervalued when they talk to scrappy voice agents, atleast make an effort to use good ones, so it tries to speak like humans
I think AI is here to stay, but AI customer support really sometimes become a hurdle than a help. Businesses should look into when the customers get frustated and add human touch.
It seems that low quality human customer support, or rather, no support at all, has been normal for some time. I don't enjoy talking to an AI, but if they are constructed to be efficiently helpful, I welcome it, grudgingly.
Almost every “call tree” I’ve needed to navigate has seemed designed to prevent customer escalation to a human at all costs, despite not meeting my needs as a caller. In this way they are user hostile, and I expect the new technology to be used similarly to much greater effect.
That being said, if the technology can remove the need for human escalation, I’ll certainly change my view.
Second, the best support line is one where no one calls in. So there are strong incentives to make it hard to do so, either positive (create good self service options) or bad (hide the number, hide emails, etc).
If you don't want this future then make sure you always ask for the human and constructively provide feedback that you will penalize any supplier that relies on it.
We are already in a K shaped economy wrt call centers. Being a very frequent Hilton hotel guest, say, gets you a separate call center number, US based live human support, people who can solve your problem, etc. Not being a Hilton member gets you the 45 minute wait and a bot.
There's always a viable alternative. But not for customers who are always looking for the cheapest viable products or services.
So those customers will spend too much time and frustration with bad customer service, because they think it's outrageous to consider a competitor who is a little bit more expensive.
At $PREVIOUS_EMPLOYER, they prided themselves on having the best support operators, with deep technical prowess of their specific knowledge domain. Their L1 were better than some other business' L3s infrastructure specialists. This lasted for more than a decade, right until a merger with another business that also had their own human support department and certain "strategic opportunities" were identified. At first, some operations were consolidated, and some people were fired. Once the corporate machine saw the numbers, they just kept going. Nowadays, its almost full AI.
Meanwhile, Gemini gave me the url and explained what I needed to know in one reply. Problem solved. Same question to ChatGPT gave me the correct answer as well. I bet Claude or Grok would have also found the correct answer.
This is where things take a turn: Google.com/ddg/bing failed. amazon's own search failed. Its worth knowing that the good results from the chatbots pointed at Amazon's web.
Based on this one example, I could begin to think document style search engines are dead.
Presumably Amazon were trying to put me into their little Voice AI microcosm. They'd obviously (?) record my voice (yes) of me getting increasingly annoyed (likely) at some Voice Chat Bot (likely) and then they'd apologize (?) for not knowing (probably?).
Is that the goal? Waste 30 minutes of my time for what could have been answered by a chat bot but not a web search?
The underlying situation seems even weirder than simply trying to cut costs. I think Amazon has cut so many corners they have forgotten what shape they are trying to create.
I have only one answer so far: Their customer support has no idea how to navigate their own website and this is by design. This is NOT the customer support people at fault. Let that all sink in: Someone got a bonus for this service design. This is somehow optimal according to Amazon. Yet this style of service design will only get worse for paying customers. It will be much worse for non-customers.
Progress!
I don't think this is restricted to Amazon either. I think this is industry wide. The in-page chat bots are usually just as broken.
Today, I get ai support which fails to resolve my problem until a human steps in.
ai support is the new elevator music
There are commercial lists of people who have histories of using customer support.
Companies use them to determine economically rational levels of resources to commit when servicing customers.
1. Native understanding of most languages
2. Full knowledge of everything in the company's knowledge base
3. Is never rude, annoyed, lazy
4. Can work 24/7
Elevenlabs using a completely incapable voice agent for sales outreach made me wonder if they were actively trying to dissuade customers.
It feels like the sort of thing a disinterested engineer implemented to appease a manager fixated on the shiny.
Give it 12-18 months and for routine support flows the gap between a well built voice agent and a human will be hard to notice. Complex or emotional cases are where humans stay, which is fine, that’s where they add value anyway.
Support agents are a brand’s first line. It hope they invest is good agentic support to protect their brand. Time will tell I guess.
Something that was table stakes in the past would now be seen as crazy for a company to spend money on
That being said, if the technology can remove the need for human escalation, I’ll certainly change my view.