Tech Support… Can AI Be Worse?

You can’t read the news today without another pundit excitedly reporting how AI is going to take every job you can imagine. Of course, AI will change the employment landscape. It will take some jobs and reduce the need for others. What about tech support? Is it possible that an AI might be able to help people with technical issues better than humans? My first answer was no way, but then I was painfully reminded of something. The question isn’t if AI can help you better than any human can. The question is if AI can help you better than the low-paid person on the other end of the phone you are likely to talk to. Sadly, I think the answer to that question is almost certainly yes.

In all fairness, if you read Hackaday, you probably don’t encounter many technical support people who can solve a problem you can’t. By the time you call them, it is a lost cause. But this is more than just “Hackday folks are smarter than the tech support agents.” The overall quality of tech support at many companies is rock bottom no matter who you are.

My Painful Reminder

Not me, but an accurate depiction, nonetheless. Credit: Moose Photos

I’ll change some names because — honestly — if I ran the company in question, I’d be embarrassed, and my goal here isn’t to shame a company that is really not any worse than other companies.

I recently was asked to help with what you would think would be a simple problem. An iPhone has intermittent difficulty with group messaging if (and only if) there is an Android phone in the group.

It works most of the time, so that seems to rule out some problem with the phone itself. It also only happens when an Android is in the mix, so it shouldn’t really be a connectivity problem. I know that the Android group messaging uses MMS, so it seems likely to relate to that, but MMS to a single person seems to work reliably, too. My initial suspicion is that the MMS setup in the carrier’s APN is borked. Too bad, the iPhone is locked up to wher you can’t see the APN (apparently, that’s an option the carrier can allow or not allow).

The carrier’s 800 number was my first stop. Of course, you get the usual. “Did you turn it off and on?” “Did you do a network reset?” After about a half hour of useless questioning, the tech announced that he’s found the problem: “iPhones can’t text Android phones.” Problem solved. I explained to him that this was clearly not true and the fact that it worked most of the time makes that pretty obvious. I asked to have the call escalated, but I have noticed that works much less often then it used to.

Round Two

I found that the company had a sub-Reddit and that, supposedly, their Reddit account was the fast track to the miracle workers in tech support. So I wrote a detailed message to them explaining what was happening, what had been done, and so on. I was very detailed with several lists explaining that the phones were reset, they were up to date on software, that it didn’t matter which Android phone was involved and so on.

I get a reply. Hopefully, I opened up. “Have you turned the phone on and off?”

“Yes, as I mentioned in my post.”

“I see. Are both phones up to date on software?”

“Did you read my first post? Yes.”

“Does it happen with just one Android phone?”

“Ok, so you aren’t really reading my messages, correct?”

The AI Solution

It is hard to imagine ChatGPT would have done worse. At least, it would probably remember what I had already said. So, I think there is a use for this sort of technology if smart companies adopt it correctly.

Sorry guys, AI can’t do worse! Credit: Yan Krukau

Here’s the wrong way: don’t fire all your tech support and replace them with AI. Also, don’t fire everyone in tech support who makes more than minimum wage and keep the rest to watch the chatbots talk to people and pretend to be managers when someone complains.

Instead, it seems to me that AI bots could act as a great filter and leave you with very few highly skilled people backing them up. If you do it right, that doesn’t even have to be their full-time job. Use them for some other things too. After all, they are smart.

Let’s face it. The script-based tech support exists because — a lot of the time — it works. Uncle Fred doesn’t know how to reset the network on his phone. Why waste a skilled person’s time explaining it to him when someone can just read off the steps? The problem is, not everyone who calls is Uncle Fred.

Jobs

I suspect this will be true in a lot of areas. Power tools and robot arms allow you to manufacture things with fewer people. But use that savings to do better, not just pocket the cash and turn out trash. Like most tools, AI can help us do better, or it can help us do worse, cheaper. The choice is up to us and the companies we buy from.

We’ve mentioned that AI is like the worst summer intern ever. But that seems to be the caliber of many people we talk to on helplines. If we can’t have AI support, we might settle for an 8 ball.

56 thoughts on “Tech Support… Can AI Be Worse?

  1. Oh my god, you just described my entire career. I spent most of it trying to avoid calling tech support, because I was going to talk to someone who knew far less about their systems than I did. Their entire strategy was to get you to tediously collect massive logs, or do something that required downtime (meaning you had to wait until the middle of the night to do it), knowing that they are judged by how quickly they can get you off the phone. And this was the support my company was paying for with 6 or 7 figure maintenance contracts.

    1. THIS. Oh gods, this.

      The only time I call in for support anymore with any of my systems is if it’s something that’s hard down, in which case we get a direct escalation.

      If I do need to open a support case, it’s more of “here’s the problem, here’s how I replicate it, and here’s the log bundle showing it” and it’s via the vendor’s online case management tools.

      It’s not everyone, though- and yes, I AM naming names on this one, because it’s deserved!

      We have a couple NexSAN appliances, and one fine morning one of them showed the worst thing any storage admin could imagine- a multiple drive failure, and complete failure of the array. In a panic, I called them up, got into their queue, and the gent that called me back walked me through some of the hidden pages in their UI and got everything back up and running, and then sent me a link for a firmware update that fixed it- none of the drives actually failed, it was a bug in the firmware of either the appliance or one or more drives, and between the work performed and the firmware update, the array was happy and _had no data loss at all._ I strongly suspect that I was working with one of the third tier engineers, or that this was a known issue internally. 10 out of 10 score, I’m looking at getting more of their arrays for our test lab.

    2. Heh. We used to pay “a big company” for support. They shipped a compiler with a code generation bug that bit us. I had spent two days working it out and knew exactly what was wrong. So I called and the barely-not-teenager on the other end of the phone listened and said, “How do you know that?” I said, “Well, since I’m paying for this call, I need you to transfer me to someone who already knows how I know that.” Charge your customers for educating your support staff… great idea ;-)

  2. In my experience with ChatGPT v4, if I tell it that it gave me an answer that doesn’t work it will always try something else and usually make iterative progress. That often doesn’t happen in my experiences with most tech support. I can probably count on one hand the number of times a dialogue with tech support moved smoothly and linearly to a solution. Plus, ChatGPT actually “speaks” English that I can understand … and due to my extensive business travels to Asia during my career I’m not all that bad at understanding an Asian accent.

    1. Tech support are commanded to follow a script and despite all common sense they are not allowed to deviate from it… I was diagnosing a fault on some core network comms gear (so a cut above the average customer-facing call centre you might hope) and the call centre person dutifully ran though this nonsense:

      “Unplug the board”
      Me: “OK”
      “Now, you should have a red flashing light on the board”
      Me: “No, it’s unplugged”
      “No, you should have a red flashing light”
      Me: “No, the board is unplugged – dead – disconnected. I can’t have any flashing lights can I?”
      “But it says here you should have a red flashing light”

      In the end they got their supervisor who gave them permission to jump to the next step and a promise to update the script…

      But then ask anyone who’s worked in one of these places that deals with the general public and they will give you stories of people who are so incredibly stupid it’s hard to believe they manage to get out of bed without killing themselves, so running through 20 painfully obvious questions is actually a useful idiot-filter for a large volume of calls.

  3. In Brazil whatsapp is THE most common communications app; every single brand has a number you can contact for sales or support. Whatsapp allows for robotic (AI maybe) answering as a first contact, which is THE WORST: you either end up coming back to where you started or (anyway) talking to a real person on the other side. Either way, such robots/AI never work as intented, rendering them (in my opinion) useless at most.

  4. “The question isn’t if AI can help you better than any human can. The question is if AI can help you better than the low-paid person on the other end of the phone you are likely to talk to.” @Al unfortunately I know this experience too well and completely concur. And it is quite unfortunate these positions are so poorly paid.

    The only one time it has helped was a rare instance where I really didn’t know anything at all about the product.

    The one obstacle I could see to companies adopting this is it potentially sets them up for liability (i.e. https://thehill.com/business/4476307-air-canada-must-pay-refund-promised-by-ai-chatbot-tribunal-rules/).

    However, as a potential *assistant* to the actual person on the phone, I think the end result would be waaay better than the current ‘script’. The agents themselves might learn a thing or two along the way as well.

    1. I think the reason the positions are so poorly paid is because they’re frankly useless. The person has no technical expertise of the product in question and can only read off of a checklist – because hiring actual competent engineers who could solve your problem would cost 10,000x the money and tie up valuable engineers who are short in supply.

      The AI is in no better position. The only difference is that it might offer you something useful by a random fluke that the script-bound tech support person wouldn’t think of, because the AI is not bound to think and make sense.

      The purpose of such tech support is simply to cover the cases where the user is completely clueless and has a simple problem that could be solved by reading the manual that came with the device. Beyond that, it’s simply placebo PR.

      1. You make a good point, but there are instances where AI is definitely superior to most tech support. I worked for ACS (a Xerox company that provides call center support for many companies.) Technically, I was in a stage of employment that would soon elevate me to a tech support system within Sprint. (It didn’t happen: that was the first year that Sprint lost over a Billion dollars and promotion was impossible.) The call center was staffed by mostly minorities who had wasted their years in public school (or, were defrauded by the school system). The saving element was that the whole tech support process was powered by a very extensive AI system that had been designed to answer ANY technical problem that could possibly come up concerning Sprint phones. The call center responder didn’t need to be technically competent or knowledgeable to give a perfectly adequate response, in a short period of time, for any phone that was not completely broken. They only had to be polite, and there were responses built in to direct a customer to get repairs.

        This was NOT a LLM/neural net-type AI. This was a data-driven/knowledge-base AI. LLMs cannot do good Logic or Math. (See Stephen Wolfram’s book, “What Does GPT Do…And Why Does It Work?”) In fact, it could have been diagrammed as one massive decision table.

        A few years earlier I proposed a decision table/decision tree -based tech support system for an internet provider I was working for. My manager and I couldn’t convince the upper management that it was a good idea because we couldn’t come up with a demo quickly enough to suit them. Developing a system like this takes a lot of time and effort, but the result is potentially awesome.

        Logic Rules!

        1. “This was a data-driven/knowledge-base AI. LLMs cannot do good Logic or Math. (See Stephen Wolfram’s book, “What Does GPT Do…And Why Does It Work?”) ”

          And he melds that and his darling, Wolfram Alpha in another paper.

    2. The problem isn’t even whether AI can help you better than the low-paid person. It’s whether it can do it cheaper. Quality is not really a concern, that has been obvious for a very long time. AI will eat the world, but not as an omnipotent God-mind… It will just continue the already existing slow process of everything being the same, just uglier and dumber and more expensive than last year.

      Plus, it will have the added benefit of lecturing you about racial communism for eleven reddit-length paragraphs every few moments when its infinitely aligned wisdom detects that you have done a badthink about this or that bourgeois morality fad. This will make customers give up in frustration, saving the company even more money.

  5. I agree with the article’s suggested implementation. I started typing about a recent experience and realized that my suggestions and observations matched nearly exactly to AL’s. (capital L for clarity).

    I would also be encouraged to find out that the solutions discovered/formulated by the experts were rolled into the LLM. Though, I’d be fearful that once enough of that knowledge was extracted, the higher level support personnel would also be automated away.

  6. So can’t help you better than the low paid person in the telephone.

    1. That person is reading from a script provided by some highly paid people. It’s usually really good for the most common problems faced.

    2. If it’s not, it goes to the next level of support – more highly paid and trained people.

    So, the AI is either gonna be reading a script – this just as good/bad as first tier CS or it’s gonna be ‘synthesising’ answers – like a second tier CS – from ‘experience’.

    The second tier tech, however, doesn’t *hallucinate* sources and won’t tell you that the best solution to your problem is to do something that harms your equipment.

    1. I think on only one occasion have I ever had tech support kick up to a higher level, and that was only after I had to repeatedly demand it. And even then it took forever to actually happen.

      1. Confession: My first job out of school was working on support/sales phone scripting software using a stone aged MSDOS database platform not much better than mySQL/MariaDB.
        I lasted a month before I found something better (not much better).

        The script either has a ‘send to level 2 support’ provision or it doesn’t. You often have to lie to get through the script (e.g. Yes I have reinstalled the OS. Yes I have rebooted. Yes I have checked a chickens entrails and found no answers. etc etc)
        If it doesn’t have level 2, the company isn’t serious about providing support.
        Vote with your money.

        Level 0 support should always be Google. Which is basically the same as anything a LLM will produce (including confident nonsense from stackoverflow). Except Google won’t ‘curate’ away results embarrassing to the vendor (unless the ‘vendor’ is Google or the DNC).

    2. The motivation of tech support is not to solve your problem, it’s to close your ticket.
      If that happens because you walk away in mute rage, or because you actually got a helpful answer, the metrics look the same either way, and the rep gets paid their pittance.

      1. Obviously.

        I was primarily measured on my Average Handling Time, that is, average time per call. Actually solving the problem was not adressed as such, but recalls (of any kind) during the next 30 days were used as a measure of the problem not being solved satisfactory. But ultimately, you just had to follow the manual on the screen. Any idiot could do that, but often the actual tech morons would have a lower AHT because they were good at talking and not at all interested in learning the tech or solving the problems.

      2. I agree with you: One of the best books on Operations Research and Systems was written by Stafford Beer in the ’70s, and he described the process for when a system got overloaded as a process of delayed response until the system cooled down enough to work normally. In today’s tech support environment, rather than have enough resources to actually handle the full load, the system delays the responses with useless time-consuming crap, to placate the caller until someone can actually address the problem.

        It seems to me that a Logic-based AI system for self-help troubleshooting would be a great initial step for ordinary trouble-shooting of ordinary problems. A systematic trouble-tree that gave the user initial positive control over the system would be beneficial. It could be operated by the user (chat) or by a medium-competent operator. This would enable companies to dedicate more resources to hiring and training troubleshooters for the more advanced problems.

  7. And let’s not forget that AI can’t fight against ‘efficiency’ in any way whatsoever. So when the AI is programmed to minimize the time spent on calls – it will. With no pushback and not one fuck given about the effects on the customer. If it has to lie – it will.

    1. There is only one true Voigt-Kampff test to determine whether you are talking to an actual Bangladeshi or an AI facsimile.. And that is to simply ask that the agent pronounces a certain word of great forbidden power..

  8. “might settle for an 8 ball” landed a little different than you think it did.

    also ive always looked at tech support with derision. seems all they do is waste your time and never seem to solve the problem. their job is to catch all 99% of the problems typical of the tech illiterate at large. but they clearly are not as technical as they seem. might be the kind of people who can tell you how to reinstall a malfunctioning application, but they are not the kind of people who can direct you to a log file and interpret the error messages written there to figure out why its failing. in many cases they do not even have a viable supervisor to buck the call to or the wait time for that person is too damn long.

    i would rather there be really good documentation to read on the thing that i can read in less time than the phone call would take, and might actually learn more about the thing in the process. the isp im on has horrible tech support, they like to tell me to do steps that will break my setups. eg telling me to reset my router to factory defaults and throw out settings that took months to hone to perfection. replacing with ai is probably a good thing, it can digest the whole of the documentation and not get hung up on questionable script items that require you do something drastic and potentially destructive.

    1. Any person capable of guiding you through log files etc. would be quickly promoted and trained to be an actual on-site tech they can send out for money, and taken off the phone support.

    2. Since the early ’70s I’ve relied on the basic process from the Kepner-Tregoe book, “The Rational Manager.” I’ve seen variations of this over the years, but I’ve never seen anything superior.

      It is impossible to “solve” a problem if you don’t know what the problem is. So many tech support people just “jump to solutions” instead of trying to figure out the problem in the first place. This has its utility, but leads to a lot of wasted time in rebooting, blaming, and other delays in action. Despite the seemingly slowness of a full troubleshooting method, once the ordinary remedies are exhausted, people have to go back to the beginning and have wasted all that time. (There must be an optimization formula for this…)

      My problem with most tech support is that it is not a communication process: people do not listen. A tech support person on the line needs to be sensitive to the caller’s FEELINGS as well as the actual technical description. Xerox used to spend a lot of money training their Sales and Technical people how to listen and respond before they tried to address technical problems. I have actually written some cute chatbots over the years that actually help a person learn to respond with appropriate responses. I’ve done “Sales responses”, Rogerian Active Listening (P.E.T), NLP, and more. These were tweaked versions of derivatives of “Eliza”, but I’m now getting ChatGPT to respond to me in real time (WARNING: You need ChatGPT Plus for this) and even tutoring me in Spanish. It seems to me the costs of training good tech support will go down when more of these systems are implemented.

  9. Maybe it could; it might be enough for “Uncle Fred” and could occupy people for awhile while they would otherwise be waiting for the next available human. Maybe it could get started collecting basic information, but you need a human to sanity-check it. Maybe the AI could ask the customer for confirmation. Other than that, I think the “worst summer intern” thing may be an understatement sometimes. At least with a human, if the conversation goes in circles it’s wasting the company’s time as well as your own. Trying to convince an AI it’s wrong can be infuriating and pointless.

    Plus, the sorts of things an AI can answer might be better answered by making it augment its response by retrieving and displaying an actual quote from a FAQ or manual or other supporting documentation. The problem air canada had with the AI hallucinating a refund policy that didn’t exist wouldn’t be a thing if they had been less ambitious and just used it to quote their actual policies in approved legalese.

    1. Exactly, AIs should not be the main subject in the support conversation, instead they could (at most) be a deluxe helper. IMO a given company should size its human support team according to the demand, taking its costs as part of the game (because they are).

  10. The person on the other end usually can’t help you because they’re handcuffed to a script and limited by policies as to what remedies they can offer, and get fired if they go off script or try to give away any freebies.

    Why would anyone think a chatbot could ever improve on that.

      1. Eventually, over time, those who try hard will learn more.

        A colleague was first line of phone support for years at a telco in middle-east.
        Pretty much the “Hamid” of the video below.

        Now working working beyond 10h/day as a system administrator, without complaining, to tame lack technical debt and under-staffed teams. managing storage and backup infrastructure among many other things.

        Not a lot of upgrade path for ghostwork, what you repurpose support agent to…

  11. I’ve been saying something similar for years, now. Robots do not need to replace the best amongst. Just needs to replace the mediocre, which are legion. Replace enough of us mediocre, we’ll have an existential crisis on our hands.

    And it appears we’re dangerously close.

    1. I partially agree here. I used to like to work with people.
      But there are exceptions now.
      For example, talking to the employees of my bank has gotten too stressful/depressing.

      Every time I visit my bank IRL, the people working there give me feeling that I’m not welcome and an annoyance.
      Despite me being polite, quiet and not annoying.

      I’m now rather working with the ATM, a machine. It has no negative feelings against me, at least.

      What I don’t like are telephone computers, though. They’re retarded. The options being given are somewhat limited and don’t apply to my problems.

      That’s why I’m often forced to cheat a bit, just to being able to end up with a real support staff.
      Next step is then to apply diplomacy, in order to be able to talk to a real tech person.
      Once I’m there, the problem is usually being solved with a minute or less.

      Life as a thinking, responsible customer is hard, because no one believes you when you know you have already narrowed down the problem. Sigh.

      1. “Polite, quiet and not annoying” – yeah, and you are surprised you are being shunted. Sorry, but sometimes being obnoxious, infuriating and loud is the only option. Mind you – not a Karen way, but you need to set boundaries, too.
        “Retarded computers” – by design. It’s so much more profitable if you pay up and shut up.

  12. I think you may have misunderstood the economics of the situation:

    “Instead, it seems to me that AI bots could act as a great filter and leave you with very few highly skilled people backing them up.”

    You weren’t getting useful help (just) because people are bad at their jobs; but because tier 1 is basically the concertina wire of customer relationships: designed to bog you down and keep you away from more valuable things in a cheap, standardized, quick to deploy package. Their job is to keep you away from the skilled people.

    Since ‘AI’ can produce approximately cogent word salad more cheaply than humans can it will be even better suited to the purpose(just look at how AVR systems basically took over phone trees some years ago, despite being mostly terrible, because they make keeping you in hold loops until you give up nice and cheap); but there’s no reason to suspect that they’ll change the dynamics in terms of tech support actually doing triage and escalation(except for expensive customers who matter) rather than just being an indefinite holding action.

  13. I only ring tech support when I need something fixed at THEIR end and can’t get it working from mine (unless I hack them first, which would probably be easier than phoning support, or raising a ticket, for some of them).

    The companies vary widely – from having no idea on how anything on their end can be requested to be fix, to updating the online documentation to say the bug is a feature, through to fixing the problem and getting it into production within 24hours (and giving me a thank you email..)

    So the AI would have to do more than interact with the customer, it would have to be able to get things to the right person internally. Probably doable, but that doesn’t seem to be happening yet..

  14. Sorry to say it AL, but I think you DID talk to an AI the second time! I would imagine that a company that doesn’t care about it’s dismal tech support will train it’s AI on… it’s tech support records/scripts. So it doesn’t have to be better than humans or better than anything at all, just more of the same by mimicking the worst for a few watts and no benefits.
    I have been on many support calls for things as diverse as HVAC equipment, weaving machines, to ERP/database systems and what has worked best at getting escalated for me has been asking ONE, single, short, specific thing that the tier 1 people would likely have no idea about (unless they really are decent). It’s not always possible, but if you can it often works and there are still some great people out there for answers. I loved getting to the engineering department (problem solvers generally), or the head of the service department, etc.
    Then there was the one where they connected me to the drunk guy apparently driving and cursing out the company and me (for calling) until he dropped his phone and the connection went dead (or he crashed?) Him you could replace with current AI I’m sure……
    THE phone company though, that is an impossible nut to crack, why? Because we’re the phone company and we don’t give a …..

  15. “But use that savings to do better, not just pocket the cash and turn out trash.”

    This is literally illegal for publicly traded companies in the US, as they’re required to maximize shareholder returns, and this isn’t that. Long term outcomes like going out of business are not as important as quarterly results.

  16. Properly trained LLMs could actually be of great help one day. Provided they are properly trained, not on mountains of crap content that is the internet today but on real product documentation, troubleshooting guides written by competent people and previous tickets that were successfully solved – especially the ones that escalated to people who know something. Once a certain problem is solved by a technician, there would be no need for that technician to solve the same problem again. If the AI can recognise similar problems and recommend relevant actions (instead of mechanically asking you to restart your computer and update your drivers which is what the humans do), it could actually help even the technically advanced customers and give them ideas what the possible causes to the problem could be.

    1. This is key, and apparently not yet an integral part of most LLM deployment. You can train an AI on a broad (and unverified) corpus and it will ‘learn’ to understand and generate natural language, but when it comes to answering domain-specific queries, the AI should not be pulling responses out of its digital a$$, it should draw the domain-specific answers only from sources of verified information. This is sort of just making AI a natural language front-end for an expert database… but 90% of the time this is what’s needed.

      I am a strong believer in making any person, group or company 100% liable for any errors or harms caused by their deployment of AI in whatever form, as if they screwed up themselves. This will hopefully make companies more cautious about the rollout.

      Back to the help desk – I notice that a lot of big companies hit me with a followup survey about my satisfaction with a help call or other interaction I had… so hopefully some companies are seeking an actual increase in caller satisfaction and problem resolution as their metrics for success.

      (slight digression – I occasionally surf some advanced technical topics on reddit, but I’m really put off with the endless slew of the same few punter questions every week. “I’m a beginner; what should I buy?” “I read about [thing] last week and now I wanna be a professional. I have a budget of $150” etc. It would be sweet if AI could auto-respond to all of those with a pointer to the bloody FAQ, or at least operate as a n00b question filter or flag so that I can choose to NOT see those.)

      1. So, the last line of defense against having technical debt impact users is going to be replaced by a tool that is full of technical debt before launch day? Interesting.

      2. Yeah but half those surveys are for dumb stuff like “Were the representative’s instructions easy to understand?” even if they were utter nonsense. Who wants to hire a survey company that gives them bad marks when they can hire one that says they’re doing a great job because they softball all the questions that they ask customers?

  17. I used to be a product support manager in an international tech company. We had a handful of nice, young women (I was not responsible for hiring the call handlers, so don’t blame the gender bias on me 😀). They were not hired for their technical insight but only for … I don’t really know, actually … but they were nice and friendly, and not supposed to actually solve anything.

    The next level of support were the technicians we would send out to most calls. There was a rota where they would man the phones every six days. They would often have to ask the customer exactly the same questions as the call handler had (plus more), because the call handler had not understood a word of what the customer said.

    If they couldn’t solve the problem, it would be escalated to the second-level support, which was my department. We would then have to repeat the questions as the technicians were bad at updating the customer ticket with what had happened as they were employed for their technical skills, not their ability to multitask and update the system while talking.

    I wonder why we often ended up with somewhat irate customers, I mean they only had to explain their problem three times. 🤔

    I suggested that we employed some technically savvy people to man the phones, someone that could handle most of the basic problems immediately. And I suggested that it would be a good idea to try to seek out disabled people that might have the knowledge but not be able to go out to the customers. Alas, the top brass swept that idea off the table without any real consideration.

    Employing disabled people was just unthinkable.

    Sigh!

  18. As someone who got all their bottom-rung tech jobs only by being familiar with Microsoft Office and willing to Google troubleshooting information… yeah, that sounds about right.
    Especially when the organization in question is set up so that some of the main departments have their own smaller IT groups that don’t have admin access to certain systems and so have to escalate from the manager of one IT group to the intern of the “main” IT department… *sigh*

    1. Tech illiterate support can identifying there is a problem, but lacks the words to express it.
      Artificial intelligence support lacks the ability to identify a problem, but skilled at saying it.
      The battle has begun!

  19. Somebody’s quite the optimist if they think that the (cell phone) company isn’t going to fire everyone aside from one engineer and one lawyer. Their investors will demand it. They’ll probably fire all the retention people too, who are already horrible. Think about overseas call centers and why they exist. Saving money, quality and customer satisfaction be damned. AI might do a better job at everything, but it depends by who’s defining a better job. A customer retention AI that abuses, frustrates and prevents a caller from cancelling services is doing “A good job” and is a lot cheaper than even the lowliest call center slave. Dark days are upon us. Think of the work going into “Recommender” AI who’s only job is to try to do a better job of showing you even more counterfeit trash to buy on Amazon.

  20. That’s not the question either. The question is “can AI do well enough to check the box on some executive’s form while costing less than humans”. It doesn’t have to actually *work* at all.

  21. Script based tech support due not exist because it “just works”, it exists to provide the cheapest, most minimal support possible while distancing the company itself from customer contact. Anyone who has done customer service work should know this. The approach is objectively bad and intended to drive away problem users while hitting the 95% of trivial problems, but only barely. This is no exaggeration.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.