The tools, the mistakes, and the guardrails, from a CEO who runs agents on his own company first.
AI agents have moved from a talking point to a job description at GiaSpace. As part of the company’s shift to MSP 3.0, agents now handle routine work across the service operation, and Founder and CEO Rob Giannini has strong opinions about how to do that without breaking your tools or your people.
So we asked him everything: where AI actually helps an IT department, what should never be automated, and what he says to businesses that are skeptical of all of it. The condensed version ran as a press release earlier this month; this is the whole conversation.
On the Record With Rob
Q1: Where does AI make the biggest difference in an IT department right now: helpdesk, onboarding, monitoring, something else?
Documentation is the big one, but it is not the only place:
- Documentation. “AI is taking the burden out of IT documentation; we are able to document a specific request or a client runbook within Claude and Notion in no time at all.”
- Onboarding. “We can use all the meeting notes to capture the client’s sensitive issues that they want to make sure are not neglected or missed in the process.”
- Troubleshooting. “With MCPs [the Model Context Protocol, which connects AI directly to the systems a team manages], we are able to work in a single pane to troubleshoot by asking AI to check NinjaOne and ConnectWise to see what is causing the issues and give suggestions.”
- Fixes. “Having AI that can run PowerShell and scripting to fix an issue is mind-blowing.”
Q2: What is the difference between an IT department that uses AI tools and one that runs AI agents? Why does that difference matter?
“AI agents run scheduled tasks without user input, while an AI tool is more of a one-off request: a chat to discuss an item or draft something. At GiaSpace, we have several agents running behind the scenes gathering data from 3CX, Slack, email, and ConnectWise to identify any service gaps, and one agent ships a report to me like a service manager would. Agents are going to take on the work that no one wants to do, and if they are crafted properly, they will run with very little intervention.”
Q3: What is the first thing a company should automate, and what should it never hand to AI?
- Automate first: “The time-consuming tasks that would give the team back time, and that the agent can have access to perform.”
- Hold back: “Sensitive work that requires exact accuracy without human oversight.”
Q4: How much faster can response times get?
“Much faster response and closure times. In the past, a senior engineer might work a specific server performance issue for days to identify the actual cause. Today you can run highly customizable troubleshooting tools to identify and resolve it almost instantly. We are also seeing that our newer hires have an AI senior engineer over their shoulder when they need one. The tech still needs to know the fundamentals; otherwise the AI can lead them down a rabbit hole.”
Q5: What is the biggest mistake companies make when they bring AI into IT?
“They do not have a clear plan for what it should do, so they use it as a Swiss Army knife for everything. And if they do not put rails in place, techs who tend to be lazy will use AI to do their job instead of using it to enhance their output.”
Q6: How do you keep the human touch, the “Friendly” in Fast, Friendly, First-Time Fix, when agents are handling tickets?
“We tell our techs that the moment we stop picking up the phone and maintaining relationships with our clients, we can all be replaced with AI. AI cannot give that personal human touch, and we have to remind ourselves to use it to enhance our services, not replace humans.”
Q7: What is the biggest mistake companies make when they start building agents?
“No real planning up front, and where are the agents going to run? We use Docker Desktop and create all of our agent systems on a dedicated machine. Having an agent created locally on a personal workstation is a risk; if that machine is off, so are your agents.”
Q8: What would you say to businesses that are skeptical of all of this?
“Do a small proof of concept before jumping into the deep end. The worst you can do is waste a lot of time trying things. But if you plan it properly and educate your people on how AI works, you can reap the benefits right away.”
The One Thing Agents Cannot Do
| Agents Take | People Keep |
|---|---|
| Scheduled tasks and behind-the-scenes data gathering | The phone call and the client relationship |
| Documentation and client runbooks | Judgment on sensitive, accuracy-critical work |
| Routine troubleshooting and scripted fixes | The fundamentals that catch AI’s mistakes |
| The work no one wants to do | The plan for what runs, and where |
GiaSpace runs on that split internally, and it is the same rule behind the AI consulting, managed IT, and cybersecurity it delivers to small and midsize businesses across Florida.
Ready to Put AI Agents to Work in Your Business?
See how AI consulting works for your business: www.giaspace.com/service/ai-consulting-services
Schedule a 15-minute AI assessment call to discuss your biggest challenges and identify automation opportunities: [Schedule Your Call]
