When people talk about AI in customer support, they usually imagine a chatbot answering simple questions. But in reality, the most interesting implementations go much deeper than that.
A good example is Hosting.com – a global hosting company with more than 900 employees and around a million support conversations every month. Their approach shows that AI in support is not just about replacing agents. It is about building an ecosystem of tools that help both customers and support teams work faster and better.
In a recent podcast conversation hosted by Konrad Keck, the guest was Panos Kesisis, COO of Hosting.com. During the discussion, Panos explained how their AI systems work behind the scenes, what problems they solve, and why data matters more than the AI model itself.
As he put it during the conversation: “It doesn’t matter how smart the LLM model is. It doesn’t matter if it’s Gemini or OpenAI. You need data.“
Customer support is still the core of the hosting business
In web hosting, infrastructure matters. Performance matters. But support is often the real differentiator. Customers rarely switch hosting providers because of a slightly slower server. They switch because their website stopped working and nobody helped them quickly enough.
During the conversation, Panos described the rule very simply. Customers want two things – a fast website and great customer support. Everything else is secondary.
That is exactly why AI became such an important topic for Hosting.com. When models like ChatGPT started showing how well they can generate text, it became obvious that this technology could support customer service operations. But instead of deploying a simple chatbot and calling it a day, the company built several AI systems working together.
The chatbot that solves up to 30% of customer questions
The most visible part of their AI system is the chatbot. It is not a third-party tool. The entire solution was built internally by their engineering team. Today the chatbot resolves roughly 25 to 30 percent of customer inquiries without human involvement.
These are usually simple requests. Customers often ask how to pay an invoice, how to reset a password, or how to change basic account details. Those tasks do not require a human agent. Instead of waiting in a support queue, the customer gets an instant answer. The result is simple but powerful. Customers solve their problems faster and support teams can focus on the cases that actually require technical expertise.
As Panos explained, the goal of the chatbot was never to replace support agents. The goal was to remove friction from the simplest interactions.
AI routing between multiple language models
One interesting detail about their system is that the chatbot is not tied to a single AI model. Instead, the first model acts as a router. When a customer starts typing a message, the system analyzes the intent even before the message is sent. Based on the type of question, it decides which AI model should generate the answer.
For example, a technical question might be routed to one model, while a sales-related question might go to another.
Hosting.com currently uses several models including Gemini, OpenAI models, and others. The router decides which one is best suited for the specific request. This approach improves response quality while also helping optimize costs.
The chatbot can perform real actions inside the hosting platform
The system does not only answer questions. It can also execute actions directly inside the hosting platform. The bot is connected to hundreds of APIs and internal systems. When a user asks for something like a password reset or a DNS change, the AI can trigger the operation automatically.
Of course, security is critical in this scenario. The system verifies whether the user is authenticated and usually requires confirmation before executing any change. The goal is to avoid mistakes while keeping the process fast.
AI that helps support agents troubleshoot problems
Even with a powerful chatbot, many cases still require human agents. Hosting issues can become complicated. Email delivery problems, DNS configuration issues, or server errors often require deeper troubleshooting. That is why Hosting.com also built an internal AI system that supports their agents during conversations.
The system reads support chats in real time and suggests possible troubleshooting steps. If a customer reports an issue with email delivery, for example, the agent can pass the error message to the system. The AI analyzes the information and suggests possible solutions.
This makes the work of support agents faster and reduces the time needed to diagnose technical issues.
AI quality control for every conversation
Another interesting use case is quality monitoring. With around one million interactions per month, it is impossible for managers to manually review every conversation. AI helps automate that process.
Every interaction is analyzed according to predefined criteria such as communication skills, empathy, and language quality. The system assigns a score and provides feedback to agents after their shifts. This allows support teams to improve continuously without requiring manual review of every interaction.
Many companies focus on choosing the best AI model. But according to Panos, the real challenge is something else. AI systems require large amounts of structured data and high-quality documentation to work well. Without that context, even the most advanced model will struggle to produce useful answers.
That is why Hosting.com invested heavily in building strong knowledge bases and maintaining large internal datasets that their AI systems can use.
AI in support is about augmentation, not replacement
One theme that appeared several times during the conversation is that AI is not meant to replace people. Instead, it is meant to support them. AI handles repetitive tasks, analyzes conversations, suggests troubleshooting steps, and monitors quality. But complex problems still require human expertise. This combination allows companies like Hosting.com to scale support operations while still maintaining a high level of customer service.
AI in customer support is still evolving. Many companies are experimenting with different tools and workflows. But the example of Hosting.com shows that successful implementations rarely rely on a single tool. Instead, they build an ecosystem where chatbots, automation, internal assistants, knowledge bases, and analytics systems work together.
In other words, AI in support is not just a chatbot. It is an entire infrastructure designed to make support faster, smarter, and easier for everyone involved.
This article was created based on the webhosting.today Podcast.
Kamil Kołosowski
Author of this post.