Human Customer Service - The New Differentiator

Do you want a chatbot with no comprehension of the question you're asking to be your customer service rep? I don't. Not for me, and not for any business I have an interest in.

The experience is terrible.

Incorrect answers. Hallucinations. Wild goose chases. Even gaslighting. A chatbot will happily tell you to eat rocks, or to open a gmail account. If bad responses make people go away, it may be incentivized to give deliberate bad responses to close tickets. Not because it's intentionally deceptive. But because it's a random text generator. At any time, it could say literally anything.

I think this is going to become a bigger differentiator between businesses than people are thinking right now. When given an option, if a space isn't just totally monopolized by bad actors, human customers will choose companies with human employees.

Customer service, I would argue, is not a space where a ML model is ever going to be a good idea. It's well documented at this point that ML models don't lie, they don't "hallucinate", they simply don't know the difference between when they generate text that is correct, incorrect, or in outer space. The error rates are a baked in part of the technology. Customer service is not a place where you I want to see these kinds of errors.

When your customer is contacting a customer service agent, your relationship with that customer is already at risk. Unless you are, say, Comcast, and have monopolized service in their municipality. Or perhaps you are AT&T and have duped the customer into a very disadvantaged contract so the cost of cancelling service is more burdensome than spending hours trying to get a real person on the phone.

But for most businesses, we aren't monopolies. Customers have other options. I think as a whole, Americans at least, are already sick as hell of useless or worse AI being forced on us by greedy executives with monopoly power.

I'm not an anti AI luddite. The tech has some genuine uses. Our county is betting our entire financial system on the snake oil salesman pretending a useful product is right around the corner. Just use this broken one for a while, and eventually by magic it will stop being broken.

Please. Don't do this to yourself. Don't do it to me. Don't do it to your fellow man. In the long run, what's good for your customers is supposed to be the greedy play. So do right by your customers.

A Real World Example

Bad validation on a broken web form was preventing me from buying a movie ticket recently. I reached out to support. Instead of letting dev know the website was broken, the AI agent told me to switch email providers.

This is just the kind of support question AI will fail spectacularly at: A question it hasn't seen answered 1,000 times by human agents on prior support tickets.

Me:

A form on your website is broken. I can't access the account for movie tickets I just bought. It is demanding my birthday to continue, but not allowing me to enter it because of a coding error. Please fix.

Support:

I reset your password to this shared bad password (e.g. Password123) please try again

Me:

Thanks. I am still required to submit my birthday before using the account, but the form is still broken. Changing my password to this insecure one doesn't help.

Support:

This happens to people who use your email provider. Please open a gmail account and use that for our website instead. Your email provider blocks password reset emails.

Me:

It's bullshit you're asking me to open a gmail account because your birthday field is broken on your website. You should let your tech folks know there is a problem, instead of gaslighting me. If they don't know about it, they can't fix it. I guess I'm done patronizing your movie theater, if I can't buy tickets.

Support:

The most AI generated apology follow up ever, like I just told ChatGPT it was incorrect about something.

Several hours after that last response, I received a password reset email from their system. I assume it took hours before a human processed the support ticket after the AI Agent gave up on it. And they chose to close the ticket by sending me a password reset.

I don't know if they fixed their form. Because I'm not going to visit that movie theater again.

Who Would Use Such a Broken Support Bot Instead of Real People?

Absentee owners.

This previously fantastic independent theater chain began to fail during Covid. They were purchased by private equity firm Altamont Capital. The PE firm fucked their workers so bad they unionized. Complaints included low pay, inconsistent schedules, on the job injuries, understaffing and union busting. Under allegations of union busting, Altamont Capital sold the theater chain to Sony.

If you're surprised Sony, a movie studio, is allowed to own a chain of movie theaters, so am I. The FTC recently decided to kill the Paramount Decree. That rule prevented studios from owning theaters because the obvious severe conflict of interest.

I suspect Sony doesn't care about the employees or customers of this theater chain any more than Altamont Capital did. The conflict of interest is the point. Controlling distribution is a movie studio's monopoly wet dream.

That's who uses support chatbots. Corporate raiders. Union busters. Private Equity. Absentee owners.

Let those of us who are not evil people not follow in those evil footsteps.