How Implementing AI Thoughtfully Could Have Saved a 20+ Year Client

I just got off the phone with my cellular carrier, and I will never use them again, for anything. 

I called them in June to ask about my options for international use, after a few attempts, I got a clear story and price. I then got back from a family trip to Europe and my bill is 3x the price described in June. I haven’t seen a phone bill that was 4 digits since landlines and long distance, so that’s a fun blast from the past?

When I called to dispute my bill, it will shock you to know I had a less than amazing experience.

Routing seemed to work pretty well. A friendly AI asked why I wanted to call and was able to parse my natural language request. It then asked a bunch of identification questions and was able to understand a 15 digit account number, a mumbled business name, and even asked for and seemingly verified my pin code. Great, we’ve just eliminated a bunch of frustrating back and forth typically done through challenging accents and poor headset connections. 

Except, no. We haven’t.

After waiting 9-12 minutes, I am now connected with Evangeline, who can answer questions about my bill - but wait, first she wants to know the exact same information the AI just verified from me. Literally every question again. I understand there are always implementation and compliance challenges on rolling out new technology, but this seems like a remarkable waste of my time and their money. Why have the AI if we’re just going to do the same exact dance with a human after? 

Eventually I got to explain my situation and had to go through the predictable rounds of avoidance most reps are taught at these call centers:

1. These are valid charges.
   They may be valid based on what you’re looking at, but they’re not what the rep told me would happen in June, nor do they match the very accessible page on your website describing the plan. 

2. Yes, but you have a ‘business’ account and those are commercial plans.
   Cool, as exciting as it is to know I get to pay more and get less service as a business, again that should have been explained to me in June when your people told me otherwise. You record calls, look it up. 

3. We can’t change your plan now.
   Okay? Again, not the deal discussed. 

4. Here’s a detailed rundown of the charges.
   Yes, I understand them - I’m telling you they are not what was expressed to me when I called to specifically get pricing on this in June. 

5. I asked my Senior Representative and they are saying these are valid charges. What we can do is offer you a payment plan so you don’t have to pay this entire bill at once.
   No thank you. I’m not interested in a payment plan for a bill I shouldn’t pay. To be perfectly direct, if you insist on charging me 3x what you told me the cost would be 60 days ago, I will pay my bill in full, I will pay off any devices on my account, I will take my business elsewhere as quickly as I can. I will never use your services for anything again in the future. I’m not trying to be difficult Evangeline, but I spent my time reaching out to you all in June, and you’ve changed the deal on me. To me this feels like an affront to my basic dignity and it’s unacceptable for our relationship. I’m prepared to shut everything down and I’d like to speak with someone who can save this account. Do you have a button to hit for that?

6. Let me speak with my manager. 
   Great. 

7. My manager says we can change some of these lines to the reduced rate you mentioned, we can offer you this partial credit.
   Well it’s nice to hear that you all understand this is wrong, but I don’t think calling this a ‘credit’ makes sense. I want to pay what you told me I would pay in June, not more. You told me X, now you’re billing me Z. Offering Y isn’t a compromise that makes me feel good about our relationship. 

(As an aside, getting what you need from call center folk is an art. My technique is to bounce back and forth between being super friendly and apologetic that they have to be dealing with this issue, with then being coherently pissed off. You have to show them you’re mad, but also give them the hope that you’re not completely insane and the relationship is eventually recoverable.)

The end result is I spent about 40 minutes on the phone and I got about half of what they were trying to charge thrown out. They kept telling me it was a “credit” and I kept pointing out they were actually asking me to spend 50% more than what they had originally said back in June, so that’s not really a “credit,” but I did end up taking it because we were reaching an impasse and my blood pressure was ready for the call to end. 

I am also going to move forward with dumping them. Moving cell carriers is certainly a time consuming exercise and not something I wanted to add to my list, but I’m afraid it has to happen. There’s no way I can look at this and feel like I was treated the way they’d want to be treated (or the way I would treat people.) That’s a sad opportunity lost for this company. I’ve been using them for easily two decades or more, and I make a fair amount of expensive decisions around IT and telecom infrastructure.

For all the talk of AI and ‘delivering delightful customer experiences’, this organization seems to somehow have lost the basic sense that any smaller business who actually cared about their customers would have. Through a boundless desire to optimize processes and reduce costs they have created a system that lets a loyal customer with a lifetime value in the 6 figures feel fleeced for a couple hundred bucks. 

Let’s examine what they’ve used AI for, what they apparently have not, and how they might have done better:

1. Call routing & authentication.
   Great, this feels like a no-brainer in 2024 and while no one loves telling a robot what street you grew up on or your favorite teacher’s last name, most of us have done it enough to not have it dramatically impact our frustration level when calling customer service.
   That said, I remain puzzled that we had to jump through the exact same information when a human support rep eventually showed up. This strikes me as likely a side effect of this company having very siloed chiefdoms of responsibility and a political bureaucracy. Whatever group implemented the AI identification didn’t impress the people running the account billing department and some policy decision has been made that “We don’t trust that, ask again.” With a more open development culture at that company perhaps a dedicated authentication team could be rolled out that actually listened to the concerns and needs of the billing department around authentication and delivered a solution that worked well across the multiple departments you’ll end up talking to in these situations. 

2. Call recording and summaries.
   Every call I make “may be recorded for training purposes.” Data storage and voice to text transcription are both relatively cheap these days. Why not have an AI make a short summary of each call and ask the support person to verify it’s accurate before automatically storing it to the CRM notes? It’s a big ask to expect each call rep to make meaningful notes on their own, particularly when call volume is often a key performance indicator. If someone had been able to look at the summaries of of the calls I made in June and realize “oh crap, yeah one of our support reps DID tell this guy that the consumer plan applied to him” it would have been much easier for a manager to accept that they had to honor that deal, and maybe even retrain the support rep who was incorrect back in June. 

3. Computing account value.
   I’ve used this company since the 1990s and my last name has never changed. I’ve purchased any number of cell phones through them, I’ve purchased internet, land lines - you name it. In those years I can’t remember calling with a complaint. I also don’t remember ever missing a payment. There’s AI that could look at my billing history and rank my account with a score that a customer rep could see that should make them believe I’m not lying about the misdirection in June to save a buck. 

4. Better KPIs.
   The person I’m talking to genuinely wants to help me, and they also want to do well at work. What you measure is what people will work to impact. Many times this company’s representatives have tried to help me by selling me something unrelated that might bring the price down. At one point I was offered to get a 5th line I didn’t need at all because it would bring my monthly price down by more than $50. This is insanity.
   
   I can see the appeal of measuring any revenue a customer support department might bring in, as traditionally those are cost centers, but might there not also be a number to value my continued relationship and satisfaction with? NPS has big drawbacks, but there are more nuanced customer experience related questions you can track. (For what it’s worth, NPS does happen to already be -18 for this company… so, there still might be something to it. ;)
   
   AI could be used to improve sentiment analysis by looking at the data that’s already been collected. These scores tend to provide better value the higher the response rate is. The natural language processing they’re using to capture and ignore my account information could be used to improve score collection to increase the participation loop. 

5. Account recovery.
   It’s a shame there isn’t an AI trolling through call center recordings now identifying accounts like mine that are in jeopardy. 

I wish large companies stopped seeing AI as a way to replace customer service with a chat bot. Instead I wish they were looking at it as a set of power tools that can make it easier for their customer success representatives to deliver great experiences to their customers at scale. 

I’ll be curious to see what happens when I’ve got my ducks in a row and call back to close everything, I'll update this blog post if they actually let me fire them. ;)

If you're interested in working with people aren't just impressed with what technology can do, but also are thoughtful enough to actually solve useful problems with it, reach out - we'd love to help.