Klarna is a Swedish e-shopping site that I had never heard of until yesterday. I remain slightly unsure of what they actually do, but they claim to handle 2 million transactions per day for half a million merchants. This means they employ a lot of customer service agents in their call center.
At least, they used to:
Klarna today announced its AI assistant powered by OpenAI. Now live globally for 1 month, the numbers speak for themselves:
The AI assistant has had 2.3 million conversations, two-thirds of Klarna’s customer service chats. It is doing the equivalent work of 700 full-time agents...and communicates in more than 35 languages. It’s estimated to drive a $40 million USD in profit improvement to Klarna in 2024.
....Klarna's AI Assistant offers real-time updates on your outstanding balances and upcoming payment schedules, ensuring you never miss a Klarna payment. It also provides a clear understanding of your purchase power, explaining your spending limits and the reasons behind them, empowering you to make informed and confident shopping choices.
For now, Klarna says the displaced customer service agents will be reassigned to help other customers. But obviously this won't last forever:
In the longer term, as more companies adopt these technologies, we believe society needs to consider the impact. While it may be a positive impact for society as a whole, we need to consider the implications for the individuals affected.
Welcome to the future. It's only going to get better/worse from here on out.
Is this AI? Or just more sophisticated algorithms? I guess the latter.
Seriously, the term AI should never have been invented. It makes us expect a radical break with existing technology. But that is not what it really is, namely more and more sophisticated algorithms that use more and more electronic (and energy) resources.
At least call it II (imitation intelligence).
"It makes us expect a radical break with existing technology. "
Exactly right. These are just more complex rube-goldberg machines; that is, they are still Turin machines.
That's gibberish.
Umm...and you say that because you know...anything? Do you know what a Turing machine is?
AI is just 'sophisticated algorithms'.
There is no way for it to be otherwise, unless it's 'magic'.
(Sorry, paywalled article: https://www.ft.com/content/c1f6d948-3dde-405f-924c-09cc0dcf8c84)
That's a pretty apt term. Turing, back in the day, proposed a test, which he imagined was the threshold for AI. I doubt that he actually meant that it would somehow equate with human intelligence.
Every time I worry about AI eliminating jobs, I go visit a grocery store. The echoes of “place the item in the bag” and other error messages in the self checkout area suggests that customers themselves are the source of many problems retailers have, and AI isn’t going to be able to fix that. The AI will still need lots of employees to deal with the many customers that simply suck at dealing with computers. (Anecdote: yesterday I went to the movies and saw 2 people struggling to buy a ticket and dispense a soda from touch screens. The inability of people to use technology is very real.)
"Place the item in the bag" is not an error message. People, including me -- a software/firmware engineer -- struggle to use technology because it is so horridly incompetently designed.
Also, there's no "intelligence", artificial or otherwise, involved. The Voice will tell you "place the item in the bag" for the tenth item as well as the first, after you've clearly figured out what to do.
That's not intelligence. That's persistence.
Exactly! Every time I shop at the same supermarket, the woman in the machine asks me if I brought my own bag. It should know by now that I always bring my own bag.
Agree, every one of these systems that I've used is very badly designed (like a lot of other "convenience" consumer tech). And according to an article that ran recently there's intense consumer resistance against them, plus they're very expensive, need attendants, and seem to have encouraged pilfering and "shrinkage." So stores are apparently re-thinking that investment, and hooray for that. I avoid them where I can.
Also I think Salamander's right that this isn't remotely AI, just primitive (and stupid) automation not fundamentally different from an ATM, and it's automation where the different implementations underline in their own different ways just how primitive they are.
The most sophisticated element about these systems was the sales job their makers' reps did on supermarket-chain management.
No, the self-checkout yells at you for stupid things.
There's no reason for the bagging thing to complain if I use my own bag.
It really depends upon where you go. The self-checkout machines at the local grocery store are awful. On the other hand, I've never had any problems with the ones at Target.
I prefer using self-checkout, because I'm a misanthropic introvert who would like to get out of the store without having to talk to anyone.
They say that regardless
Is it any different or better than the standard phone tree?
"Please say or type in your account number"...
Really. If the customer checking their outstanding balance is thing, the answer is to log in to the account and just look. Using chat, or even worse calling, seems an insane way to do this.
A few years ago I had to cancel my NY Times subscription. Every month, I would receive an email informing me that they could not charge my debit card. Every month, I would look at it in disbelief: somewhere, from the bowels of a poorly culled out database, they would extract my old debit card that had expired, and which I hadn't used in a couple of years. The exact same thing happened with Netflix. For whatever reason it defied fixing.
I would attempt to have a conversation with the customer service monkey. However, their script only included two sentences. Literally! Would AI -- whatever it is -- be any worse than customer service reps who had no understanding of technology and complete inability to grasp what was going on. Furthermore, even if they did understand what I was trying to tell them, it was not within their power to do anything about it.
AI failure was on the backend, resurrecting my old debit card due to some faulty heuristics. It truly makes no difference whether we would have AI on the front end. Dumb looks are still free, whether they come by the way of AI or live people.
I recently got a new card. This turns out to be a good way to filter for unneeded subscriptions. I let the Hulu subscription lapse. I would renew it if anything in the family complained, but so far no one has even noticed. Hulu stopped emailing me about it after a month or two.
In my case, I wanted to hold on to the subscription, but it just became too annoying to deal with this every month. The great mystery, which remains unsolved: where did they dig up my old card. This may happen to you, too, when you decide to renew your subscription.
I have something to look forward to!
This all depends on how well the AI does things. Customer service these days is heavily scripted, which makes it very machine like anyway. Usually, if I am to the point of calling customer service, I have to wade through 20 minutes of telling machines and/or people that no, I already tried rebooting and that didn't help. And here is my account information ... again.
Eventually, I will get to a human being that A) knows something and B) has the ability to do something about my problem or tell me something valuable. This is never quick.
One reason that it's never quick, and usually painful is that it's expensive to field calls, and they would rather you didn't. Maybe having a computer system handle it would be better in that there would not be the incentive to make people avoid calling but an incentive to solve their issue quickly?
Honestly, an AI could quite likely do this better. And I prefer doing text chat over a phone call anyway.
Can't speak for this particular company, but of the two dozen or so companies I've had to interact in the past year which have switched to virtual assistants to answer questions, provide info, etc. Every single one of them has been absolute crap. Human assistants weren't always helpful, but they never batted 0.000 like all these pretend "AI" assistants do.
"In the longer term, as more companies adopt these technologies, we believe society needs to consider the impact. While it may be a positive impact for society as a whole, we need to consider the implications for the individuals affected."
The idea behind this is completely backwards. An efficient AI assistant that makes the customer happy would have a positive impact on that individual alone, not society at large. However, putting scores of people out of a job doesn't just have a negative impact on the affected individual, but on society as a whole.
1000 happy customers for ACME means squat for society. 1000 former ACME employees now unemployed and disgruntled can bring down the economy of an entire region when you account for the domino effect of all those people losing their job.
Implementation is always a crapshoot. If I'm forced to call, very rarely have they given the power to the computer to fix whatever it was.
But I have had real people fail to have the tools, too. Some are just frustrating, like when Amazon refused to send me a PS5 to replace the one stolen from their truck on its way to the post office. And they didn't refund me until the end, where the lady refused to call this situation a bug or a problem and she was doing a favor refunding me. Ugh.
When people are put out of a job by technology, that is hard for them in the short run, but in the long run it frees up labor for other uses, which is good for society. This is how the economy grows and how our standard of living improves.
You're gonna miss those Pakistani customer service reps.
Indeed. They may be hard to understand at times, but they have human intelligence, flexibility, and typically, a lot of technical knowledge and ability to fix things.
I can do without the Pakistani/Indian spammers, though. No, I wasn't recently involved in an auto accident. No, my business doesn't need a big big loan. No, I'm not looking for a Medicare Advantage plan. And if you claim to be from "the Department of Social Security", I will say bye-bye.
The Medicare Advantage calls seem to be petering off, so there is that.
Air Canada just lost a court case where it used an AI in its customer service and the AI hallucinated a policy. The Canadian court held the airline to the hallucination. Companies will need to retask customer service agents to QA agents to monitor non-routine calls. Callers will also respond by starting their calls with "Speak with a human" to skip the AI menu.
You don't?
Please press 1 to accept our terms of service…
Thinking that some court decision is going to shift the balance of power is truly the triumph of hope over experience.
There still are one-off things AI "Assistants" can't handle. Simple stuff, yes-- but that's stuff we can mostly do online anyway. Just this past weekend I ran into an issue I had to talk to a live person about: My phone was stolen and I wanted the shipping/tracking info for my new phone (I had coverage on the phone). I could only get that online, or from the chatbot if I could verify my identity with a texted verification code-- and I didn't have a phone of course. So had to talk to a person who could verify me by other means.
Airlines excepted*, if I am calling customer “service”
1: I have a complex problem that cannot be resolved by their website…or AI
2: I am in a pissy mood
3: The last thing I want are more menus or “virtual assistants”.
* Airlines are all-hands-on-deck when things go to hell. I’m trying the online chat and waiting on hold for a phone call, while waiting in the customer service line. It’s war and I press all available avenues simultaneously
This sounds pretty basic and not all that different from a typical “press 1 for your balance” etc. But I look forward to it getting better. Honestly, the bar is pretty low to surpass the typical human customer service rep. Invariably I can tell they don’t know the answers to any of my questions and are just following a written flow chart to answer the most common questions. Anything that requires actual product knowledge, let alone creative problem solving, is beyond them. At least AI could be programmed for proper grammar and clear pronunciation.
In fairness there’s one exception: Apple. Every time I call them I speak to someone who actually really knows the product. And if I have a novel problem they are able to think about it and work with me to figure out a solution on the fly. Pretty impressive in today’s world.
My stepdad recently bought a car. He corresponded with a customer service rep who told him that they'd be expecting him when he came to the dealership to buy the car. When he got there... no one was expecting him, and the car he wanted was being detailed and was unavailable. The rep had been an AI that, like all currently existing chatbots, knows how to string words together but has no idea what it is actually talking about.
He did get it sorted out and bought the car the following week, but he wasted a lot of time because of this fake customer service. Of course, we've known for years that a lot of companies don't care about customer service and do it as cheaply as they can get away with, but this still represents a new low.
My health insurance "help" line is now using a robot who keeps you on the phone forever telling you to speak full sentences because he understands them. He doesn't understand anything unless it's some common question that they've programed into him. He also manages to put in advertisements. It's horrible.
Just what we need. Customer service that does this:
https://arstechnica.com/information-technology/2024/02/chatgpt-alarms-users-by-spitting-out-shakespearean-nonsense-and-rambling/
This is just a way to outsource the outsourcers, replacing a barely functioning Filipino who barely speaks English with a barely functioning bot prone to bursting into streams of gibberish. But it's free and much better at fending off irate customers who want to speak to the supervisor.
I don't know about you, but most of the time I can already do what most automated customer service things can do for me, like check my balance. I only call it when I CAN'T figure out something and usually the automatic systems are worthless on that. My #1 goal in calling these systems is to find the fastest way to talk to a human being as possible. And since companies have gotten very good at making their customer service people work very hard, they usually give me good service.
Unless there is some quantum leap at play here, I suspect the Swedish customers are hating this.
This is the same thing that's been true of Microsoft Windows "help" for ages. It only tells you how to do what you can already figure out from the menus. If you want to do something that's not obvious, it provides absolutely no help at all.
It's only going to get better/worse from here on out.
At least in this one area, I'm going with "better." I think the biggest driver of shitty customer service is low staffing levels. Being told "Your wait time is approximately seventeen minutes" to speak with a human is practically a cliche of modern life. Eliminating wait times would be a huge plus.
Cutting costs and quality by slashing customer service jobs and switching to a semi-functional automated system is not exactly 'The Future'. Its also The Past and The Present.
At least it will be easier to make out what they are saying than the almost incomprehensible Indian accents I have to wade through.