Follow Slashdot blog updates by subscribing to our blog RSS feed

 


Forgot your password?
Close
typodupeerror
×
AI Technology

The World's Call Center Capital Is Gripped by AI Fever - and Fear (bloomberg.com) 61

Posted by msmash from the closer-look dept.
The Philippines' $38 billion outsourcing industry faces a seismic shift as AI tools threaten to displace hundreds of thousands of jobs. Major players are rapidly deploying AI "copilots" to handle tasks like summarizing customer interactions and providing real-time assistance to human agents, Bloomberg reports. Industry experts estimate up to 300,000 business process outsourcing (BPO) jobs could be lost to AI in the next five years, according to outsourcing advisory firm Avasant.

However, the firm also projects AI could create 100,000 new roles in areas like algorithm training and data curation. The BPO sector is crucial to the Philippine economy as the largest source of private-sector employment. The government has established an AI research center and launched training initiatives to boost workers' skills.
This discussion has been archived. No new comments can be posted.

The World's Call Center Capital Is Gripped by AI Fever - and Fear More

The World's Call Center Capital Is Gripped by AI Fever - and Fear

Comments Filter:

  • thankyousircanyoupleasetryhittingresethardersir (Score:4, Insightful)

    by Pseudonymous Powers ( 4097097 ) on Wednesday August 28, 2024 @12:27PM (#64743966)

    The only remaining hope I had was that some call center lackey was, after thirty minutes, finally gonna break and turn me over to the next tier of support, where they're very occasionally allowed to actually do anything at all. Once AI takes over, even that slim hope is gone.

    Well, to be fair, I never did really enjoy using goods and services much anyway.

    • Re: (Score:3, Interesting)

      by Anonymous Coward

      At least eventually the call taker will either get fed up and hang up (which will hose up their call stats), they will try to throw you onto another ACD (again, a call stat dig), or maybe get you to someone who has the ability to might take over.

      AI, on the other hand isn't going to be something that businesses worry about call center metrics on. It will be a magic tool, just like offshoring that is use.

      This type of stuff is what bred the Karen archetype. When businesses cheaped out on CS, making it the bi

      • Re: (Score:3)

        by sjames ( 1099 )

        Exactly this. When consumers no longer believe that a warranty will be honored or that "support" will actually help, they'll just buy direct from China so at least they get ripped off for less.

      • Re: (Score:2)

        by HiThere ( 15173 )

        People SHOULD stop paying attention to "name brands", because the name brands no longer stand behind their products.

        • Most alternatives are just the same products under a different name from the same Chinese factory, except with less quality control, and absolutely NO support or warranty.

          Choices, choices.
      • I wonder how long it will be until we have to agree to binding AI arbitration.

    • Re:thankyousircanyoupleasetryhittingresethardersir (Score:5, Insightful)

      by Gilgaron ( 575091 ) on Wednesday August 28, 2024 @01:18PM (#64744162)
      My experience with the AI first tier stuff is it fails to be helpful at all a few times before apologetically routing you to a human, so it isn't too much different.
      • The problem with 1st tier phone support (human or AI) is their functions are only what you can do yourself on the website. So you don't call them unless you don't need them, other than talking them into escalating asap.
      • My only experience with AI support so far is a live chat that purported to be powered by AI but was really just a menu tree. There was, however, no obvious way to escape the menu.

        • Re: (Score:2)

          by dfm3 ( 830843 )
          That's basically what "AI" currently is anyway, it's just a really huge tree...

          • Re: (Score:2)

            by HiThere ( 15173 )

            Unfair. They're calling a lot of that stuff AI, but that's not what AI is. The problem is, they don't want to bother investing the time it would take to train a real AI...which would probably be better than most level 1 support staff I've talked to. But a menu tree is just a fancy script, so it's the same thing exactly, except there's nobody to get his feelings hurt if you're rude to them.

            • Re: (Score:2)

              by dfm3 ( 830843 )
              to be fair, real AI doesn't exist, which is why I put it in quotes. My point is that it's all just a fancy predictive text, just with a really huge dataset. Of course these days, every company has to sound like they're on the AI bandwagon, so they train a "chatbot" with a very small database of customer service tree options and the final product sucks. That's why the typical experience is "I can't help you with that, here are some things you can ask me about... I can't help you with that, here are some thin

              • Re: (Score:2)

                by HiThere ( 15173 )

                A real AGI doesn't exist. There are lots of actual AI programs, but they aren't general, they're domain specialists. There's one that's a specialist in "Go", there's one that's a specialist in protein folding. Even ChatGPT is a real AI, but it's specialty is "sensible sounding sentences" rather than correspondence to the external reality. And "being intelligent" in a domain doesn't mean you can't be fooled. Heuristics *ARE* fallible. But it means you'll be right in most cases without needing an imposs

      • Their hope is that you get frustrated and quit 'the process'.

  • "Algorithm Training"? (Score:4, Insightful)

    by rsilvergun ( 571051 ) on Wednesday August 28, 2024 @12:32PM (#64743992)
    no what they mean is scams. There's a large number of mechanical turk scams going on right now with fake AI firms that hire cheap foreign labor to pretend to be AI long enough to scam investors.

    There might be some short term jobs sifting through data to train AI, but by their very nature they're short term.

    We have a 4th industrial revolution starting up. For anyone who knows about the social and economic upheaval the last two actually caused that's not a good thing. We have not exactly progressed much as a species. We're not ready for this.
    • Devalued labor exported to more devalued labor than then gets even more devalued after being exported again.

    • We have a 4th industrial revolution starting up. For anyone who knows about the social and economic upheaval the last two actually caused that's not a good thing. We have not exactly progressed much as a species. We're not ready for this.

      You are disposable. Who cares if you are ready or not? If you are not ready, you die. Very simple and of no consequence to the people who actually matter. This will not end well for anyone, even for the people who matter.

  • belong to AI

    AI haz cheezburger

  • Lol (Score:2)

    by jvkjvk ( 102057 )

    So, they think that the unemployed people will get jobs to improve "quality"?

    When has quality ever mattered to corporations unless it helps profit.

    Pretty sure that these supposed jobs won't help profit they will be seen as a cost center.

  • ...not in the near term
    I predict that future AI will be able to provide excellent customer service
    Unfortunately, today's AI will be even worse that the foreign moroons in outsourced call centers
    Prepare for a tsunami of suckage

    • Honestly it can't be worse than the outsourced solution in my opinion. A company's reputation is built on price, quality, and dependability. Support issues fall into the last category.

    • Re: (Score:2)

      by dfm3 ( 830843 )
      Perhaps, only if your goal from "customer service" is to get information. If you have to call support to have someone take some action on the back end that you can't fix as a user, you absolutely don't want a machine learning interface to be able to do that.

    • ...not in the near term
      I predict that future AI will be able to provide excellent customer service
      Unfortunately, today's AI will be even worse that the foreign moroons in outsourced call centers
      Prepare for a tsunami of suckage

      It won't be good in the long-term either.
      As you say, the short-term will be a tsunami of suckage in customer service.
      However, when AI gets good enough to truly deliver excellent customer service, it will also make 1-2 billion people unemployable. So get ready for another tsunami of socioeconomic suckage

      In the long-long term it will be good, because we will finally learn the answer to the Fermi Paradox. After the global wars and ELEs have finished, we won't have the crutch of all that free fossil energy to s

  • God I hate this term...

    algorithm training and data curation

    Aka "Feed the machine guy" job. Sorry, "role".

    The future seems bright the world over. Yeay AI.

  • Man, must be nice.

  • Taco Bell in my area is running an AI ordering bot at the drive through. It's absolutely TERRIBLE. You have to repeat yourself multiple times. It always gets things wrong and then seems unable to correct the item until a human finally takes over. Its ABYSMAL and they should get rid of it immediately.

    I can't wait to talk to one of these on a phone call. Truly horrifying idea.

  • Net loss of employment (Score:3)

    by Baron_Yam ( 643147 ) on Wednesday August 28, 2024 @01:05PM (#64744110)

    -200k jobs. 200k people looking for new work that doesn't exist... Or they wouldn't have been in call centers.

    It's true, eventually things will sort themselves out .. but the short term will be brutal and you might not like the long term 'sorted' state, either.

    • -200k jobs. 200k people looking for new work that doesn't exist... Or they wouldn't have been in call centers.

      It's true, eventually things will sort themselves out .. but the short term will be brutal and you might not like the long term 'sorted' state, either.

      Everyone wants the Star Trek universe where computer and robots run things and people are free to pursue things that make them happy, like flying a Starship, or writing books, or doing research work. But literally no one wants to vote to make that a reality. "Go get a job freeloader!" We need to start building the world we want, not finding reasons to maintain the status quo.

      • We need to migrate in that direction, for sure. Right now we could use a wealth tax, a reduction in the standard work week, and a UBI sufficient to prevent starving in the streets. Then we need to adjust our economic system so it isn't based on eternal growth.

        We have more than enough productivity to afford this, though admittedly there would almost certainly be an average reduction in standard of living for anyone who isn't already poor.

        • we need to adjust our economic system so it isn't based on eternal growth.

          Absolutely. But there are many people who are never satisfied with their wealth level or power level.

          there would almost certainly be an average reduction in standard of living for anyone who isn't already poor.

          There doesn't have to be. If you choose to work and pursue a career, then you can have more. If you choose to stay on UBI and write books or whatever, you will have to learn to live on less.

        • Right now we could use a wealth tax

          One that's getting some traction is the land value tax. [wikipedia.org] A carbon fee and dividend [wikipedia.org] also acts as a wealth tax, while earning revenue to help fund a UBI.

          • Right now we could use a wealth tax

            One that's getting some traction is the land value tax. [wikipedia.org] A carbon fee and dividend [wikipedia.org] also acts as a wealth tax, while earning revenue to help fund a UBI.

            An LVT is a terrible idea. I live in Texas and we have "property taxes". I have to pay $15,000 a year to rent the home I own from the government. And I don't live in an expensive neighborhood. This makes retirement very difficult for some people. What do you tell a 70 year old widow who has to pay $1500 a month in LVT out of her $2000 social security check?? BAD IDEA.

            All taxes should be consumption based. Period. We can choose not to buy a new TV. We can't choose to not pay property taxes.

            • Where I live we have a bunch of taxes... federal and provincial services are funded by income tax, but municipal services are paid from property taxes. Then we have sales taxes and a bunch of specific manufacturing taxes. Each type of tax has a different activity generating it, and different applications.

              There's nothing wrong with a property tax that funds your local roads, garbage pickup, and sewer systems. If that "70 year old widow" wasn't paying property tax, the government would get that money some

            • Re: (Score:2)

              by Ichijo ( 607641 )

              The nice thing about the LVT is that it's only assessed on land area, not floor area. This means that with a LVT, each floor of a 4-story building pays 1/4 as much tax as a 1-story building on the same plot of land.

              Paying less tax is a good thing, right?

  • Looks like Karma to me.

  • Yet more hype (Score:4, Interesting)

    by VeryFluffyBunny ( 5037285 ) on Wednesday August 28, 2024 @01:13PM (#64744144)
    I bet LLMs aren't particularly fit for this purpose & may never be (They appear to have reached practical & theoretical limits in terms of model sizes). LLMs "hallucinate" 80% of the time, they're incredibly energy intensive to run, they require high-expertise workers to configure, maintain, & keep updating the guard-rails & prompt filters. For all this trouble in an already highly efficient & heavily optimised sector, how much in savings is this actually likely to bring?
    • there's very little competition left in markets thanks to 40+ years of weak or no anti-trust law enforcement. So if every company forces you to fight through an AI customer service system to get what you need then you don't have a choice.

      You can't vote with your dollars in an oligarchy. The King doesn't care if you buy his iPhones. Welcome to Techno Feudalism
    • LLMs don't hallucinate 80% of the time in general. In certain areas yes. Can a better prompt be made? Sure. I could see LLMs having a potential as level 1 help. They can go through the same script the humans are now. Neither understand what they are doing most of the time anyways.
      • Sorry, mixed up my sentences. LLMs are only right about 80% of the time. They "hallucinate" around 20% of the time. This is on average, which varies depending on the subject matter; the less common, the more the hallucinations. Another word for what LLM vendors call hallucinations is bullshit. LLMs are very confident bullshitters. For a definition of bullshit, see: https://en.wikipedia.org/wiki/... [wikipedia.org]

    • The use cases cited: summarizing customer interactions, and assisting with suggestions, seem to me to be actual, useful things that LLMs can do. How many times have you been on a 5th call with customer service, and were told to pleas hold while the agent read through the notes? If LLM can summarize these in a couple of seconds, this will definitely be an improvement. And in my experience, LLM suggestions are at *least* as useful as human script-reader suggestions.

      • Have you spent much time reading LLM summaries of texts & then comparing them with the originals? Yeah, they do summarise but what they focus in on seems to be pretty random, often missing the more important dimensions of meaning in the texts. There's an "art" to effectively summarising texts that glorified autocomplete isn't very good at. Who knows, maybe for their purposes, if they're simple declarative, low-level texts, it'll be good enough?

        • In my experience with human call center agents, your depiction of the performance of LLMs sounds like an improvement.

    • Re: (Score:1)

      by Tablizer ( 95088 )

      I bet LLMs aren't particularly fit for this purpose & may never be (They appear to have reached practical & theoretical limits...

      I suspect we'll see a fairly big leap when somebody figures how to hook LLM's to other types of AI, such as something like Cyc.


  • If you just spend a few billion on AI all that stuff you sepnd of a few million on can go away. Just imagine the possibilities!!

    https://hbr.org/2023/11/keep-your-ai-projects-on-track
    https://www.businessinsider.com/ai-return-investment-disappointing-goldman-sachs-report-2024-6
    https://www.cognilytica.com/reasons-for-ai-project-failure-overpromising-underdelivering/

    Sir, did you hear the latest Nvidia AI accelerator provides at least 17 million times more AI perormance?? BUY BUY BUY!!!
  • Those call centers apparently overlooked that ancient mantra - Evolve or Die

  • OH NO! Not my heckin tier 1 supportirinos whose job is merely to funnel people out of the system by being slow and ineffective. Won't someone think of the integrity of the Phillipeans tier 1 customer support industry?

    Anyway

  • Try talking to the Rogers bot "Anna", it's an AI chat bot that can't answer a single question that I have ever asked it. I can understand not answering an open ended question, because you need to reason and direct more questions, but it can't answer specific, detailed, exact questions.

    Anna is actually worse then the "typical" Rogers customer service experience, and if you've called them, you know what I mean. Rogers customer isn't just bad, it's completely unusable. I recently had a problem where the

    • So Amazon has their executive escalation email address which still seems to go above and beyond regular terrible support. Even if it's bad, every tech or tech adjacent company needs a shibboleet code [xkcd.com] to be able to send a message to the right contacts.

      My local TV station had an audio problem for weeks one time. It was just bad enough to be annoying but just good enough that the average person wouldn't know where the blame lies. I eventually just started sending messages to any email address I could find a

      • That comic is hilarious!

        My favourite excuse is: "Oh, you're running Linux? we simply can't help you, just can't, the problem you're not running Windows or MacOS." I've had companies from Atlassian, Microsoft, ESRI, Rogers, Bell, Lenovo (.... they blamed a hardware design problem on Linux), and others just hear "Linux", and refuse to continue the conversation.

        Both Bell and Rogers have blamed modem issues, on the fact my computer was running Fedora. The TV box issue, a tech was over a year ago, and no

  • Most of what I'm seeing is trying to automate the maundane tasks that takes humans not interacting with humans. Grading agent performance is what I'm seeing the primary focus on. Being able to take all the data about the calls, all the recordings, and even the agent's screen interactions; and not just generate an agent's grade off that, but look at how the center is operating as a whole.

    It's just...expensive. That's not to mention this isn't just being offsourced to cloud; but it's vendor hosted. Depending

  • because most people are incapable of reading basic instructions, so, most calls can be solved, but for all other uses cases, is like a big pile of horse shit

    • I agree with you. Summarizing the case so far, is a perfect use for LLM. And AI suggestions are probably at least as helpful as what you get from today's script reader humans.

  • Let's hire them to answer our phone and filter out all uninteresting calls...
    Guess it won't be long until that job is replaced with AI.
    If you think things through, it often ends up with an AI generating content and the user using AI to filter the content. Why bother with an old medium like phone lines or email? Let the AI talk with each other directly in a vector space. Saves a few nuclear reactors.

  • Most call center workers are doing it because the hiring threshold is really, really low. They'll basically take anyone who can read from a script. Anybody who's actually good at it, either gets quickly promoted, or gets burned out. Turnover is really high in this industry. From what I can see, AI will be a good thing here.

Slashdot Top Deals

If all else fails, lower your standards.

Close