Verizon has turned to Google Cloud’s Contact Center AI to automate phone calls
theregister.com
theregister.com
I have never phoned customer support to:
- View my account details (available online)
- Find out about new plans and promos (available online)
- Pay my bill (available online)
- etc (available online)
I have only phoned customer support to:
- Negotiate a better rate (requires a human)
- Argue about excess charges on my bill (requires a human)
- Get put on a off-market plan (ie loyalty, retention, etc -- requires a human)
I guess if there really are that many people phoning support with trivial questions, a voice assistant will free up the phone line for people who actually need a human. I suspect it won't, as they'll reduce the number of support employees accordingly.
It's no coincidence that these systems make getting to those points harder and harder.
The one time I've talked to a human to cancel a service, it was Three and it was a pleasant surprise because although their retentions team did end up keeping me I got a huge discount over not only what I was paying before, but compared to the competitor I was leaving them for. I have no idea how it's even profitable to give me service at the current price, maybe it isn't and just drives somebody's retention-based bonus?
This is unfortunately not a freedom we enjoy in the US. I recently changed my car insurance provider to a much cheaper one bundled with my home insurance. I called to cancel the old one and talked to the rep for like an hour. He didn't even really try to save my account, made a few half-ass efforts but mostly just chatted. It was early in the pandemic lock down and I didn't have anything else to do anyway so I just talked to the guy for a while. A month later I notice another charge from them. Turns out he just ... didn't cancel it. I think he kept me on the phone so long for plausible deniability if his call got audited, and then just marked me down as a retained customer. When I called to cancel the second time they did actually cancel and refund me back to the original cancel date, fortunately. This was Progressive.
"I'm moving overseas for an extended period of time." (Implication: to somewhere you don't offer service)
"Where are you moving?"
"I can't really say."
Turns out, most service providers have an exclusion to their retention metrics, where "customer moving to location we don't offer service" is one of the few reasons to just process their cancellation request.
Try loading a modern web form over 3g. Try loading a modern web form on an Android <6.x device. Try loading a web form on a feature phone. Try loading a modern web form in BFE over a landline. For any transaction where you don't need a human and it's painful to use the internet walking through phone menus is a pretty good alternative. They only make it painful to speak to a human. For routine transactions they're basically like using a slow CLI where they give you the help without you asking.
They work REALLY REALLY hard to get people to use the website.
It's mostly older generations, they refuse to use a website for many of those basic things. A lot of it is an issue of trust, they just don't trust websites.
When they have to pay $0.62/message, it cuts down on the crap.
I’m a member of a profession where you’re supposed to read all of their comms. Now that they’ve switched to email, they send way more crap that I have to sift through and I can’t block them.
Being (arguably) legitimate companies, their unsubscribe links actually do work.
Then there’s the emails saying: “your bill is ready”, but you gotta login to see what it is and open a PDF.
Paper is more convenient.
And the telcos make it REALLY REALLY hard to use the damn website.
I don't trust websites. Websites from regulated industries (banking, finance, telecom) are especially slow, confusing and poorly built. Totally understand why people avoid them when they know a better alternative.
Most of the time I'm calling because my internet is down and I can't go to the website. Or I'm calling because I forgot my password for a website used once every N years.
90% of the time I'm calling, I wouldn't expect the website to have the information I need. Looking at Comcast's website I'm not inclined to look around and find out.
There are a lot of people in this world (especially older people) who are used to doing everything over the phone, and don't know how to use the internet and don't want to learn.
You are never going to encounter those people on an online forum, but they exist.
I’m worried for him as the presumption of an internet connection and some basic ability to use a website increases, especially for government services.
One example I saw several years ago: The state of Texas switched to a centralized process of issuing birth certificate copies for things like passport applications. For people like us, it’s super convenient - answer some questions, pay $30 with a credit card, paper in hand in two weeks. For him, it was going to be: get librarian to print this form, mail in with copy of driver’s license along with $80 check, wait about 8 weeks.
Needless to say, I ordered for him, explaining what information I was giving each step of the way.
I design, implement and support contact centers for a living (think Avaya, Cisco, etc..). For the past five years, we've been preparing ourselves through education and training for this omnichannel take over of voice. While other channels have taken some of the voice pie, it's not nearly as significant as those of us in this field thought it would. I've created chat bots, IVRs and everything in between for automation, and people absolutely hate it and simply want to talk to a human. Most customers push going to the website, but the problem is that a lot of people don't know how to navigate properly. To help this, a lot of customers implement co-browse into their chat solution, but that's only helped a little. We too are using Google's Dialogflow and I'll be looking to deploy my first one in a couple months - very promising compared to other solutions out there like Amazon Connects Lex Bot. Regardless, leaders and technology evangelists in this field still preach that by 2025, the overall percentage of the pie for voice will end up being about 15%. But they emphasize that those humans who make up the 15% will be the most important in the customer service/support space.
The UI starts in a Word document. Then the comms people toss colours and words into it to the point that I as a developer don’t know what parts of the system do.
Never once have norms of the web been considered in the design.
>In total, across the OECD countries, 26% of adults were unable to use a computer.
https://www.nngroup.com/articles/computer-skill-levels/
I've worked for companies with large call centers, and anecdotally I don't have a single doubt about it. To close this gap computers need to be easier to use... these people aren't even at a skill level where they can reach the web.
Saying to your boss that the VP of something is incompetent and made a crappy site will get you fired. Saying there's this new Tech at google coded by god himself will get you a new cost-center and a promotion.
For the company, it will turn out as nice as it turned out to all those places renting mainframe contracts well into the 2000s.
What data points is this prediction based on?
> You demand a human. The human is told what to say by the bot
This has always been something I've never understood about customer support. In particular, chat support is bad about this, but I've had it happen on phone support too. Prior to getting connected to the human, the bot wants to take down a short description of your problem. That's fine, but the conversation often goes like this:
> Bot: Please write a short description of your problem.
> Me: Problem I am having.
> Bot: One moment while I connect you.
> (a small eternity later)
> Support: Hi. How can I help you?"
> Ctrl+C problem, Ctrl+V problem*
> Support: I see you are <verbatim repeats problem; with no paraphrasing and no demonstration of understanding, I have no idea if the agent "gets it">. Is that correct?
> Me: :shrug:^W Yes.
… like, why did I ever type that description? Here, a computer (I don't even know if it qualifies as "AI" given how rudimentary of a task it is performing) appears to have failed at the simplest of tasks. And this happens all. the. time.
Phone bots will often "authenticate" me with various credentials, and then connect me to a human who will immediately authenticate me with the same credentials. Whose time was saved?
The techno optimists (execs running these companies) often end up adding workloads because they screw up the human based systems that are already in place.
If they are ESL they may be trying to match what you wrote to their internal knowledge system and asking you to confirm.
The nurse enters the room, asks me why I'm there and types the answers into EMR. She leaves...
The doctor walks in and says "so, what brings you in today?"
It's either a huge redundancy or a useful way of sending off your message to join millions of others, with the goal of training the system to improve and speed up the automated service.
I had a problem last year with a major credit card processing bank, where they decided I needed a new card/number, but apparently were unable to actually ship said card to me for an extended period of time. During which, I was locked out of the online/automated phone/etc system because I didn't know the new card/account number (and couldn't activate it). They literally wouldn't take my money, and it took mashing the '0' button in the automated system, which kept requesting my account number, for about 20 minutes before it decided I was allowed to speak to a human. That human located literally on the other side of the planet from me, also didn't know what to do about it, and it took being escalated through three levels of customer support before I got to someone who mysteriously had the power to ship me a new card within a day or two so I could pay my bill, activate the new card before traveling overseas/etc a week later. Thats hardly the first time I've had a problem with some bank/etc making some sweeping/automated change and not thinking through all the problems it will cause to various "odd ball" customers.
Anway, automated customer support is a joke unless your just calling to check your balance or some other activity you should be able to pull off with the web page/etc. At the end of the day, you need people who can understand the system, and solve the problems created by unthinking middle managers making broad decisions.
Companies that are known to have good customer support are the ones that have given their first line support people the ability to make financial decisions that could hurt the company if misused. You simply can't (yet) do that with an automated system.
One thing you can try is the 'lost or stolen card' number rather than the general request line. They will get you taken care of most usually. Also, it's likely they already mailed your new card, so your card was indeed lost, only by them. I think this happened to me once.
If I remembered anything from the education it's this:
You have to address the customer's emotional needs before addressing the customer's business need.
If you don't you end up with an unhappy customer who had his problem solved. I don't see an AI satisfying my emotional needs any time soon.
My emotional needs are that they fix their damn service.
I would hope the Google Cloud speech recognition (used by Verizon) of the fake voice from Google Assistant (used by the phone customer) would have a 100% accuracy rate.
Of course, maybe Google could encode an alternate data stream inside the generated voice, which would enable perfect transcription.
Now I'm imagining a future where robots communicate with each other over audio channels....
Try it next time! "____! Give me a _______ human you ____! ____!!" This is often followed by a prompt click of the line transferring and sure enough you're in the queue for a real person.
Of course, remember to do the opposite with the person. Be polite and respectful with the person on the other end of the line. And I've found, they often can really help you!
A few weeks ago I started getting automated calls from “Alex at Verizon” (obviously a computer) reading a script reminding me that the equipment needs to be returned. The only problem? Sometimes it would call 5 times a day, leaving a new VM every time.
I think it should be illegal to claim to be human, and generally these messages should always be delivered by text message with an option to reply STOP.
On the plus side, I brought the equipment to a UPS store where they scanned the barcode stickers directly into their system, handed me a receipt and an hour later Verizon had emailed confirmation of receipt. I didn’t even need to pack them.
A well designed service doesn't let phone agents do anything the customer couldn't do themselves. It is literally called customer support, because they help you do the things you want to do, rather than doing it for you.
As such, their online portal offers actions typically requiring a phone call, such as phone number porting, SIM unlocking, choosing a new number/region, and cancelling your account.
Aside: I once accidentally ported the wrong number over since autofill screwed up the input box (444 from my personal 444-4444 line and 5555 from my work line 555-5555). Within 5 minutes I my phone would ring 444-5555. I feel bad for the owner of that number, because it was assigned to me without any authentication whatsoever. Really put the 2FA attacks into perspective.
The profit motive of corporations is directly in conflict with the duty to social good. As we proceed into an automated future, our labor protections are pricing humans out of the market. A great deal of unnecessary evil and pain worldwide results from this conflict.
At the scale of corporations such as Verizon, even a cost efficiency of one cent per transaction can translate into tens or hundreds of millions of dollars on the bottom line. Simple state machines with rigorously defined transitions are very efficient. Creatively handling exceptions and undefined states is a capability of human employees, and it's often very inefficient to do so. A million automated transactions could occur for every one bespoke transaction requiring ingenious human attention. Ideally, corporations would prefer not to hire people at all, but where does that leave us as a society?
Not just that, but the skill and knowledge requirements to compete with automation will continue to grow and less and less people will be able to meet that demand. Many people argue that many thought this would be the case with many technical revolutions before, but fail to admit that it's converging on a singularity. Sure jobs we didn't know would exist are being created where people can shift their work to, but the cycle for automation is becoming shorter and cheaper and sooner or later we as humans won't be able to keep up (unless you meet the high demands which are unreasonable to expect everyone to adapt to).
What kind of democratic solution do you see us humans make in the first three decades of this problem (when it really starts to hurt a lot of people)?
And what kind of realistic solution would be better?
I ask because I don't see many good solutions in my life time that will solve any of this. As you can guess I'm skeptical of UBI where people can live comfortably in the first decades due to culture and I'm from northern Europe.
This would be a great benefit to most since it creates jobs and gives us real humans to speak with. Of course, it would increase prices of things, but depending by how much (I'm sure there's a sweet spot), it may be worth it for the better customer service.
Like, if your service goes down, how are you supposed to find the new number?
I thought telecom tech supports numbers were sacrosanct.
My Comcast experience probably wouldn’t be hurt by replacing them with even a slightly malevolent AI.
Best Buy was probably the most egregious example of this.
1. Do I plug in the green wire in the green port or in the red port (They call before they have even tried to connect things).
2. My TV isn't working when they haven't pressed "Source".
3. Something is wrong, restart router.
4. My WiFi is slow when they have one router for 200 m2 in a house of concrete.
5. A few % of the costumers make up the majority of the calls because of how they are.
Turned out to be a UART issue: the 8250 would error on 9/10 packets and we needed to get a 16550 serial card. Took a very long time to get someone to determine that.
They probably lost money for a couple years on our account.
- This customer called 3 times in the past two days and we told them to reboot their router. Maybe it's something else. (I've had personal experience with this and it took over a month to get someone to replace my line from the pole to the house).
- In the case of retail, it is much more complex (returns, repairs, warranty, pricing, etc).
It turns from an "OK" experience to a very bad one pretty quick when the script doesn't solve it.
They would reach a "sweet spot" where the fee would still be affordable but dissuasive enough to cut down on the amount of stupid calls, and would actually allow them to hire less, but more skilled support advisors.
I don’t need to be told every time “Did you know...you can find us online to X, Y, and Z?”
Verizon customers are paying more because they expect more, so this move by them is disappointing.
I on the other hand use Siri all of the time when I need to be hands free or using my Watch.
Me: "I need to get a refund"
Bot: "Which order do you need refunded?"
Me: select order
Bot: "What's your reason for requesting a refund?"
Me: "Item did not arrive"
Bot: "Alright, I've credited your payment method x#### used on this order."
Fastest refund I've ever gotten-- the product never arrived at the door and usually a refund like that requires an escalation to prevent fraud; I'm assuming this system uses some sort of risk analysis score to determine if it can refund without a return.
I published that one far and wide!!!
typically, a human answers the call and is rather tentative (I'm guessing my call gets flagged due to the swearing?), and after my cheerful response back to them it becomes a pleasant interaction.
Will try the swear words, kinda did it anyway but didn't notice any improvements...
Amusingly, if you get through to a human, or demand to speak to a person, the staffer will probably just tell you what the AI wanted to say anyway: the software will, behind the scenes, provide prompts to the call-center workers.