.webp)
The world is changing fast, and AI is changing faster. At hosting.com, staying ahead of that curve isn't just a thought. It's how we operate. So when we saw the opportunity to fundamentally rethink customer support, we took it.
If you've reached out to one of our customer care teams recently, there's a good chance you were already talking to Orbi, our in-house AI Customer Success agent. And hopefully, you left that conversation with your problem solved.
Who is Orbi?
Orbi is an AI customer success agent built entirely in-house by a team of four engineers in Bangladesh, developed over the course of about a year. It's a deterministic agent, meaning it doesn't guess or improvise. It understands customer intent, securely authenticates into accounts, and takes actions on the customer's behalf with the same level of access the customer themselves would have.
What makes Orbi genuinely powerful is that combination: deep system and troubleshooting knowledge, paired with the ability to actually do things and not just answer questions.
What does Orbi do?
Orbi is the first agent that accepts every single live chat and ticket that comes into Hosting.com. Here's how a typical interaction unfolds:
1. Orbi reads the incoming message, understands the context, and identifies the customer's intent.
2. Based on that intent, Orbi summons the right subagent. Think of subagents as assistants, each expert in a particular domain: billing, website troubleshooting, cPanel support, and so on.
3. The subagent reasons through the problem, gathers all relevant information, and calls the tools, APIs, and system connections it needs to investigate.
4. Once the subagent has done its work, it hands the findings back to Orbi, who continues the conversation with the customer.
All of this happens in a matter of seconds. To the customer, it feels like a single, seamless conversation.
The results have been significant. Orbi currently handles approximately 4,000 live chats per day with a 55% resolution rate, and 2,000 tickets per day with a 30% resolution rate. Our CSAT scores are sitting at 85%, with an average live chat response time of 16 seconds and 38 seconds on tickets.
Complex troubleshooting: How Orbi manages large context workflows
Orbi isn't a simple chatbot running through a decision tree. It's designed as an agentic system that continuously reasons, acts, observes results, and then reasons again. This iterative approach is technically called a ReAct (Reason + Act) loop. Each cycle incorporates the current instructions, conversation history, available tools, newly gathered information, and results from previous actions.
To keep this efficient and economically viable at scale, we separate Orbi's context into two layers:
Static context sits near the top of every prompt. This covers things that rarely change: system behavior, safety rules, formatting requirements, general operating principles. Because this portion is nearly identical across interactions, it becomes a strong candidate for context caching, meaning the model doesn't need to reprocess it from scratch each time.
Dynamic context changes as the conversation progresses. For each customer query, Orbi injects relevant playbooks, skills, live incident data, topic-specific guidance, new messages received during processing, and the results of tool and subagent calls. This layer lets Orbi adapt to the current state of the task while staying anchored on its stable foundation.
Context caching isn't just an engineering optimization here. It's a core architectural decision that allows long-running, multi-step agentic workflows to be fast, scalable, and cost-efficient. Without it, the economics of running Orbi at this volume simply wouldn't work.
Turning support conversations into consultative sales journeys
Orbi does more than resolve issues. While it talks to a customer, it's quietly building a fuller picture of who that person is, what they're trying to achieve, and what they might need next.
That picture can include past invoices, ticket history, spending patterns, active services, geography, the page they arrived from, referral source, and competitor comparison signals. From all of this, Orbi assesses whether there's a sales opportunity and what it looks like.
Critically, Orbi follows a consultative sales approach. It doesn't push products. It runs a discovery process. The goal is to understand the customer's business, constraints, urgency, budget, technical requirements, and buying intent before ever making a recommendation. The sale comes from fit, not pressure.
Orbi's sales behavior is structured around eleven stages:
1. Prospecting: Orbi scores the customer before engaging, looking at signals like invoices, tickets, spend, geography, and landing page. This includes proactive outreach to customers who show signs of interest but haven't started a conversation yet.
2. Qualification: Orbi evaluates buying intent and decision authority, separating genuine readiness from casual curiosity.
3. Understanding the Business: Before recommending anything, Orbi learns what the customer actually does and where their pain points lie: storage, CPU, traffic patterns, compliance needs, reliability requirements. This is the heart of the consultative approach.
4. Product Mapping: With business need and spending patterns understood, Orbi commits to a specific recommendation. The product is framed as a solution to the customer's stated problem, not a generic pitch.
5. Objection Handling: When customers push back on price, timing, migration risk, or trust, Orbi treats those objections as information rather than obstacles. Each pushback reveals something useful about the customer's situation.
6. Negotiation: Negotiation only happens after the customer has settled on a product. This protects the consultative flow and prevents premature discounting before value has been established.
7. Closing: When the customer is ready, Orbi generates the invoice proactively rather than asking for permission to proceed. A customer who's already decided to buy shouldn't face unnecessary friction.
8. Delivery: Orbi checks that the customer can actually complete the purchase: invoice accuracy, currency, regional payment methods, anything that could create a dead end at the final step.
9. Retention: After the sale, Orbi monitors usage and account signals, reaching out proactively with messaging framed around value realized rather than features used.
10. Upsell and Cross-Sell: Orbi identifies expansion opportunities based on real signals: high utilization, repeated support issues, new use cases, under-provisioned infrastructure. No irrelevant pitches.
11. Voice of Customer: Orbi gathers outcome stories from satisfied customers through light follow-up. Those stories become trust proof that feeds future sales conversations.
The core discipline underpinning all of this is simple: no product pitch before Orbi understands the business, and no negotiation before the product is agreed. That discipline is what keeps the experience consultative rather than pushy, and it's what makes recommendations feel earned.
## Monitoring a quarter of a million conversations a month
Hosting.com handles close to 250,000 conversations per month across all channels. A significant portion of those are now handled by Orbi. Monitoring that volume manually, case by case would be impossible.
So we built (and yes, vibe-coded!) an automated quality monitoring system. It fetches every single conversation across all channels, evaluates each one against a defined set of criteria, scores it, and generates recommendations. Those are then automatically surfaced to our agents and managers for visibility.
The result is 100% visibility across all (11) of our offices and 500 customer care agents, in real time, with dramatically reduced manual quality control effort.
What's next for Orbi?
We're currently exploring AI voice, with the ambition of having Orbi handle phone support with high resolution rates and response speeds to match what we've achieved on chat and tickets.
Beyond voice, we're continuously improving Orbi's playbooks, subagents, and deterministic reasoning by reviewing the hundreds of thousands of conversations it handles. Our targets are a 70% resolution rate on live chat and 50% on tickets, all while keeping CSAT above 90%.
Looking further ahead, we want Orbi to be present across the entire hosting experience. Imagine being inside your hosting panel and simply asking Orbi to handle something for you, whether that's a configuration change, a migration, or a troubleshooting task. No support ticket, no waiting. Just a conversation.
This is part of a broader vision we have for what hosting infrastructure looks like in an agentic world. We believe the next generation of hosting panels will be designed for both humans and AI agents equally. A developer builds an application using AGI, asks it to get the site live, and the AGI comes to Hosting.com, talks to Orbi, and Orbi handles the provisioning, migration, and deployment automatically. No human in the loop unless one is needed.
It's an ambitious vision. But given where we are today with Orbi, it doesn't feel far off.
Rakibul Islam, the AI Systems Lead for Orbi said:
"What we love about this work is hearing customers happy at the end of a chat. Those small moments are what the team works for, and they're why we sit with real customer chats every day looking for opportunities where Orbi could have done better, and tuning how it handles that kind of case next time. The team learns from every conversation. Orbi learns from every fix. Hosting is a dynamic and varied business, with something new coming up every week. So the work is honestly never done. But every fix, every adjustment, every small improvement, is aimed at creating more of those moments where a customer leaves the chat genuinely happy with the outcome."
Customer happiness isn't a secondary concern. It's the whole point. Orbi exists to make every interaction faster, smarter, and more useful for the people on the other end of the conversation. Everything else follows from that.
