Why customer support was the right place to start
Like many companies, hosting.com began experimenting with AI several years ago.
The original goal was straightforward. Could AI help the team understand customer conversations better? Could it identify patterns, surface opportunities for improvement, and help teams respond more effectively?
Early versions of our AI customer success agent focused on analyzing support interactions and identifying trends. The results were promising enough that the team decided to take the next step and build something customers could interact with directly.
From the beginning, Panos says the focus wasn't on building an AI product for its own sake. The objective was helping customers get answers faster while maintaining the quality and accuracy they expect from support.
That sounds simple. In practice, it turns out to be a much harder problem than building a chatbot.
Why a chatbot wasn't enough
Customer questions aren't all the same.
A billing question requires different information than a technical support request. Troubleshooting a WordPress website requires different expertise than helping someone choose a hosting plan.
As the team built Orbi, they realized a single AI model trying to handle every scenario wasn't the best approach.
Instead, they created a system of specialized agents.
Today, Orbi uses 18 primary agents along with dozens of supporting agents designed for specific tasks, platforms, and hosting environments. When a customer submits a question, the system identifies the intent behind the request and routes it to the most appropriate specialist.
Different AI models are used for different types of work.
Gemini handles much of the customer-facing conversation because of its natural communication style. OpenAI helps with general knowledge questions. Claude plays a major role in troubleshooting because of its ability to work through complex technical problems step by step.
Rather than forcing one model to do everything, Orbi combines the strengths of each.
The challenge nobody sees: accuracy
One of the most interesting parts of the conversation had nothing to do with customer support workflows.
It had to do with what happens behind the scenes when AI has access to systems, data, and customer accounts.
As AI conversations become more complex, models have to process more information. When too much information accumulates, accuracy can suffer.
For a customer support system, the consequences go beyond giving someone a bad answer.
The system also needs to avoid taking incorrect actions on behalf of customers.
That challenge shaped much of Orbi's architecture.
The team built guardrails around what the system can access, what actions it can perform, and how confident it is in the answers it provides. Every response receives a confidence score. If confidence drops below an acceptable threshold, the conversation is transferred to a human team member.
According to Panos, preventing hallucinations and ensuring accurate actions has been one of the most important parts of the project.
Customers don't care how impressive the technology is if the answer is wrong.
What changed when Orbi moved into support tickets
The first version of Orbi focused on live chat, which gave the team an opportunity to test how customers interacted with AI in real-world support conversations. Many live chat interactions are relatively straightforward. Customers are often looking for help with billing, account access, domains, or product questions, and a fast, accurate answer can resolve the issue.
Support tickets presented a different challenge.
Unlike chat conversations, tickets are often centered around a problem that needs investigation. A website may be down. An email may not be reaching its destination. A recent update may have introduced an unexpected issue. In those situations, the customer isn't simply looking for information. They're looking for someone to understand what's happening, identify the likely cause, and help them reach a solution.
That distinction became a turning point for the team.
As Panos explained during the conversation, the systems and workflows that worked well for live chat weren't enough for the deeper troubleshooting required in support tickets. Solving more complex issues required Orbi to gather more context, connect to additional systems, evaluate multiple possibilities, and maintain accuracy throughout the process.
The move into tickets forced the team to rethink how Orbi worked behind the scenes. New specialized agents were introduced, the architecture evolved, and new methods were developed to help Orbi manage larger amounts of information without sacrificing reliability.
The early results have been encouraging. Within weeks of launching on support tickets, Orbi was already resolving roughly 35% of incoming tickets while maintaining customer satisfaction scores in the mid-80% range.
For customers, that means getting help faster when issues arise. For the support team, it creates more capacity to focus on the situations that benefit most from deeper technical expertise, investigation, and human judgment.
Perhaps more importantly, it demonstrated that AI can contribute in places where the goal isn't simply answering a question. With the right systems and guardrails in place, it can help navigate the process of solving a problem.
What happens when support becomes proactive?
One of the ideas Panos returned to several times during the conversation was the possibility of helping customers before they ask for help.
Today, most support interactions begin with a customer reaching out.
AI creates opportunities to recognize patterns much earlier.
A customer may own a domain but not have privacy protection enabled. A website may be approaching resource limits. Traffic trends may suggest it's time to consider a different hosting environment.
Those signals already exist.
The challenge is identifying them at the right moment and presenting useful recommendations.
Panos described this as a consultative approach rather than a sales approach. The goal is to understand what customers need and provide relevant guidance based on their situation.
That same thinking could eventually extend into the hosting panel itself.
Imagine logging in and seeing helpful observations about your website's traffic, resource usage, security posture, or growth trends. Instead of digging through reports and dashboards, customers could receive insights in plain language and decide what actions make sense for their business.
The bigger opportunity
As the conversation shifted beyond customer support, Panos shared a broader vision for how AI could help organizations work together.
His focus wasn't on replacing teams or reducing human involvement.
It was on making information easier to access.
Most organizations already have valuable data spread across dozens of systems. The challenge is finding it, interpreting it, and turning it into useful decisions.
Whether it's customer support, sales, marketing, or operations, AI has the potential to connect those systems and help teams spend less time searching for information and more time acting on it.
The technology matters.
The workflows matter more.
The companies that benefit most from AI will likely be the ones that use it to remove friction, create visibility, and help people make better decisions.
What agency owners should do next
As the conversation wrapped up, Cory asked Panos what advice he would give agency owners who know they should be exploring AI but aren't sure where to begin.
Panos's recommendation was simple: start by looking at the work itself.
Pay attention to the tasks that happen repeatedly throughout the day. Document your processes. Look for workflows that require the same steps over and over again. Those repetitive activities often reveal the clearest opportunities for automation and improvement.
That mindset is reflected in the way Orbi was built. The project didn't start with a question about which AI model to use. It started with a customer experience challenge: how do we help customers get answers faster while maintaining the quality and accuracy they expect from support?
For agencies, the lesson is similar. The most valuable AI projects are rarely driven by the newest tools. They're driven by a real business problem, a source of friction, or a process that consumes more time than it should.
Start there, then use AI to build a better experience for your team and your customers.
