The question: an AI-native CRM
Towards the end of the livestream, a viewer asked Nathan Ingram, the agency trainer here at hosting.com, whether he’s used Twenty, a free, self-hosted CRM that has premium features such as agentic AI. Nathan wasn't familiar with it, but his stance on AI-native software was very clear:
Any software like this, I'm not spending a dime on unless, from the core, it supports agent AI. I want an API, better yet an MCP. I'm actively working to replace QuickBooks with an accounting option that better supports AI out of the box, because why would I buy a product that doesn't let me use the single biggest thing I use every day? So if Twenty is an AI-first CRM, I'm interested.
That’s pretty much where the discussion on Twenty ended, but since the CRM is capable of handling agent-style work as part of its paid tiers, rather than having to open up a separate AI program to do it, it inspired us to write this blog post.
AI is a lot more than “open Claude, ask it something.” AI is no longer the destination and it’s increasingly becoming an embedded tool in software you are already using, doing work you’d otherwise have to do by hand or ask a chatbot.
AI now turns up in more places than you’d expect, and developers across categories are building it directly into the software you already use: from CRMs to marketing tools helping agencies turn websites into results. Here are a few of the most common examples.
CRM (Customer Relationship Manager)
A traditional CRM is, at its core, a structured storage for contact info, deal stages, a timeline of past interactions, all organized for a person to read through and act on.
AI changes what the software itself can do with that information. Instead of a person reading all that information and deciding what to do next, the CRM can now draft a follow-up itself, summarize where things stand, etc.
Some, like the AI-first Twenty CRM from our viewer's question above, go further and act as an agent on routine relationship-management tasks, rather than just organizing information about them.
Scheduling and coordination
Shared calendars will likely always be a thing. People need to know when things will happen and we don’t see that changing anytime soon.
However, when an AI is natively integrated into a calendar, it can help parse requests in plain language, check availability, and propose or book next steps instead of leaving that work to a human.
Booking tools like Trafft, for example, now include an AI assistant that can answer setup questions directly and accurately, cutting out a step that used to mean digging through documentation.
Search and research
Online search and research still means clicking through links, reading articles, and noting down the most important bits of information. However, with the help of AI and AI-driven search tools, that process becomes more straightforward.
AI can now synthesize across resources and reason through a question, handing back an answer instead of a starting point. It reduces the steps necessary to arrive at a conclusion.
Google's AI Overviews are the clearest example of this. Most people already know about them; they now surface directly in search results rather than requiring a click through to a separate site.
Project management
Boards and task lists usually require someone to keep an eye on them, assign them to the correct person, fill in task cards, etc.
AI can now turn a plain description into a full-blown plan of action and update task status based on what’s actually happened, rather than relying on what someone remembered to log.
Automation tools like Zapier have added this directly: describe what you want in plain language, and its AI will scaffold the workflow for you.
Customer support
Customer support has, for the most part, always meant a ticketing system, a call center, a knowledge base, and a live chat. Those typically have a human at the other end, answering requests and editing documentation.
Now, AI can triage incoming requests and work through technical troubleshooting, resolving a meaningful share of customer care requests without a human ever touching them. Tools like Thrive Desk, for example, let you embed a chatbot on your site that references your help desk documentation and can open a ticket on its own. That, naturally, frees up human agents' time for more complex tasks and requests that require greater attention. It's also an area hosting.com has been investing in directly, with our own AI-assisted customer support system.
As you can see in all these examples, the shift is toward AI becoming a capability built into the tools you already use. That's a much bigger and more interesting prospect than any single chatbot.
Orbi is a good real-world example of what “AI embedded in a product” looks like in practice, rather than in theory. We sat down with the team behind it to understand how it works.
Instead of asking one AI model to handle every kind of customer question, Orbi is built as a system of specialized agents. It uses 18 primary agents, plus a dozen supporting ones for specific tasks and platforms. In addition, it uses specific models based on what they are best at:
That kind of guardrail matters more than it might seem. In June 2026, attackers used Meta's AI support chatbot to hijack several high-profile Instagram accounts, including the Obama White House account and the U.S. Space Force Chief Master Sergeant's profile, simply by asking the bot to change the accounts' email addresses. Meta fixed the issue, but it's a clear example of what happens without the right guardrails. Orbi's confidence-score handoff is designed specifically to prevent that kind of failure: when the system isn't sure, a human steps in before anything happens.
Within weeks, Orbi has managed to move from live chat to responding to support tickets as well and is now resolving a meaningful share of customer queries while keeping customer satisfaction high.
What all of this means for agency owners
The takeaway from this blog post isn’t that you need to rush out and get an AI CRM or build a multi-agent support system.
Instead, as Nathan himself said, most useful AI applications start with a real, specific problem rather than a fascination with the technology itself: integration rather than awkward shoehorning in. That also means being honest about your existing stack. If a tool you already rely on is adding real agentic capability, that's usually a better bet than tacking on something new. It's worth asking whether a new AI-first tool is solving a genuine gap before you add it to your workflow.
That's also the advice Panos Kesisis, the COO at hosting.com, gave to agency owners during the interview we linked above: look at the work itself and pay attention to what tasks repeat every day. Document that process, since the workflows with the same repeated steps are often the best opportunities for AI to help.
And while our viewer was asking Nathan if he had used a CRM with AI capabilities, the expanded question, in our opinion, is if AI can take over a routine? Yes, it can, when you are looking at the software already sitting in your stack.
If you have any other questions about AI integration in the tools you use, AI in general, WordPress, or hosting, join every Thursday at 2 p.m. EST for Office Hours. Nathan and the community are always happy to provide answers and help.
FAQ
Is AI only useful as a chatbot like Claude or ChatGPT?
No. While chatbots are the most visible and widely discussed form of AI, many software categories, including CRMs, scheduling tools, search platforms, and customer support systems, are building AI capabilities directly into the product itself rather than requiring a separate chat interface.
Not necessarily. The most useful AI applications tend to solve a specific, real problem rather than being adopted for their own sake. It's often worth checking whether a tool you already use is adding genuine agentic capability before bringing in something new.
What is Orbi?
Orbi is hosting.com's AI-powered customer care agent. It uses a system of specialized AI agents, including different models for different tasks, to triage and resolve customer support requests, with guardrails in place to hand off to a human when confidence is low.
How can I find good opportunities for AI in my own agency's workflow?
Start by looking at tasks that repeat often. Document the process for those recurring workflows; the more consistent the steps, the better a candidate that task is for AI to help with.