I have answered these tickets since 2008.

The first queue I worked looked the way every queue still looks: email that stopped sending, DNS that had not propagated, a disk that filled overnight, a site that was down only from the owner’s side of the screen, and a customer who needed it fixed an hour ago. I have spent eighteen years with some version of that list.

In 2011 I started a hosting company of my own. One person, every role, the support inbox included. As the customers grew I hired and trained my own support staff, and learned the queue from the other side: what new techs struggle with, what a knowledge base has to cover to keep answers consistent, what you can hand off and what stays with the owner. The company kept growing, and in 2018 it sold.

A year after the sale, an established hosting company brought me in as CTO to rebuild its support from the ground up. In 2021 another US host had me do the same, and both desks still run on those systems today. Since 2023 I have been a Senior Linux System Engineer at one of the very large US hosts. I have now built the support system at three hosting companies, my own included, and seen this industry from one man to a staff of hundreds.

Somewhere in those years the complaint became a design problem. Most of what a hosting desk handles is known work. The fixes exist, hundreds of them, documented in runbooks and in the heads of tired engineers. What did not exist was a way to deliver those fixes the moment the customer asks, at 2 AM, in their own language, without waking anyone and without handing a bot the keys to the server.

I looked for that tool for years. It did not exist. It still does not, outside of this one.

So I built it the way an operator would, not the way a chatbot company would. PanelBot knows which customer is typing and what they own, because it lives inside your billing portal. It checks the actual server before it says a word. It fixes what it diagnosed only after the customer agrees, and the dangerous actions demand the kind of confirmation that cannot happen by accident. Everything lands in an audit log you can replay. The cases that need judgment, the suspensions, the disputes, the genuinely strange ones, go to your team with the diagnosis already done.

It started as my build, but it does not ship as one man’s work. Around it from the first release has been a small crew of industry veterans: engineers who have run hosting infrastructure for years, and support staff I trained myself at the desks this story comes from. Every release goes through people who know this work inside out before it goes near a customer.