Sometimes You Still Need a Human on the Other End

I spent last week migrating our payment gateway from Xxxxxt to Dxxu.

Not because of pricing. Not because of features. Because when I needed help moving from sandbox to production, nobody on their end could give it.

Here’s what happened. I had everything running fine in the sandbox environment — webhooks, callbacks, settlement flows, all tested. The next step was the production cutover. And I had questions. Nothing crazy — just the kind of specifics you need before you flip the switch on real payment flows.

I opened a ticket. Got an instant reply from a chatbot. Then an email from an AI agent that quoted the documentation I’d already read. Then another automated message saying my issue had been “escalated.”

Three days of this. No human ever replied.

The Thing About AI Support

Look, I build AI tools for a living. I run Hermes agents that automate half my workflow. I’m not anti-AI customer service.

But there’s a gap between what AI can do and what it pretends it can do.

A chatbot that routes common questions? Great. An AI agent that surfaces relevant docs? Useful. A system that masquerades as support while never escalating to a human who can actually look at your account and give you a straight answer? That’s not support. That’s a deflection machine.

What bothered me more was the thought: if this is the experience before I’m live, what happens when there’s an actual production incident? When real transactions are failing and I need someone to act fast? If they can’t help me migrate, they sure won’t help me when shit hits the fan.

Why I Moved

So I migrated. Several days of rewriting endpoints, updating webhook handlers, testing sandbox environments on a new platform.

The migration itself was fine — that’s not the story.

The story is that after the cutover, I had a question about Dxxu’s recurring billing API. I opened a ticket, expecting more of the same. Instead, a human replied within an hour. An actual person who read my question, understood the context, and gave me a specific answer.

That interaction alone was worth the migration cost.

This experience also reshaped how I think about writing about support experiences — the specific story is what resonates, not the generic lesson.

The Lesson

AI is great at scale. It’s terrible at care. And poorly designed automation costs more than you think — not just in money, but in customer trust.

When your product handles people’s money — payment gateways of all things — the support channel needs a human exit ramp. Not a chatbot that deflects until the customer gives up.

What sealed the deal for me wasn’t a feature comparison. It was the fact that before I even went live with Xxxxxt, I already felt alone. That’s not a risk worth taking with people’s payments.

Human will be human. That’s not a bug. It’s the feature AI hasn’t figured out yet.