Key Takeaways
In the span of four weeks in April, four major IT management vendors launched what they’re each calling agentic AI platforms. Kyndryl went first on April 2, SolarWinds followed on April 15. Kaseya and Auvik, which introduced its Aurora platform the last week of the month, both dropped theirs a day apart.
Now, most IT teams already have some version of AI, like Microsoft Copilot, a chatbot that makes suggestions when asked. These vendors, on the other hand, are basically saying that’s not enough to handle sophisticated IT stacks.
Just take a look at the numbers: Auvik said 48% of IT professionals say their teams spend 10 to 20 hours per week on end-user tickets, according to Auvik’s 2026 IT Trends Report.
“Today’s IT technicians are more overloaded than ever with alerts and tickets, limiting their ability to operate proactively,” said Dan Zaniewski, Auvik’s CTO.
The Training Data Is the Product
Ask any MSP the difference between a useful AI tool and one that’s not, and the answer usually comes down to what it’s actually trained on.
Kaseya’s platform, for example, is built on more than 1 billion help desk tickets, 3 exabytes of backup data, and 17 million managed endpoints. Auvik’s dataset spans 15 years of network management, 2.2 billion CLI commands, and 12 million devices.
Nothing has to be started from scratch, which is good — that’s the kind of thing that MSPs need more than just want. Hosting providers managing hundreds of separate customer environments are also often at the mercy of accidentally mixing up data with less sophisticated agents.
For Auvik specifically, that means each customer environment stays separate. “Each time the AI agents evaluate an environment, they use the latest view of the customer’s infrastructure, so recommendations reflect current infrastructure context rather than generic assumptions,” Zaniewski said.
The payoff isn’t immediate, though, he added: “Teams can also reduce escalations and preventable incidents as they use AI agents for more proactive work, such as identifying devices that need patching or replacement before they become customer-impacting problems.”
But getting there is half the problem because there still seems to be a pretty decent-sized adoption gap.
Why Is Only 5% Actually Using It?
Auvik’s research shows 67% of IT professionals say they’re optimistic about AI’s potential, but only 5% say it’s actually core to their operations. And yet, AI management is expected to become the top area of managed services investment by 2028, according to KPMG’s 2026 Managed Services Outlook.
You’re probably wondering how both can be true. The answer is that a positive outlook doesn’t guarantee a smooth transition. It’s actually where a lot of people are stuck right now, and they don’t know where to go.
Think about it: If those on the ground — like support and sales — aren’t familiar with the tooling, then they won’t know what to do. And if the people handling customer data don’t know what the AI is allowed to do, that screams liability.
The AI Optimism Gap
Source: Auvik 2026 IT Trends Report
Liability as in governance — compliance, regulations, guidelines, frameworks. The point is, you can’t just integrate AI without proving it’s safe enough to handle customer data. So, many people are trying to keep up, with 76% of IT leaders saying they have an AI policy in place, but only 42% of help desk staff say the same.
Again, it’s still a work in progress for many. And if left untreated too long, it could have serious consequences. As Zaniewski put it: “Day-to-day IT friction often comes from small unknowns rather than major outages: sorting through alerts, looking up commands, and piecing together context across tools.”
Just Remember, The Rules Are Still Being Written
People are feeling the pressure, especially with ongoing bills being pitched to the House and Senate like crazy as of late.
In March, Sen. Marsha Blackburn (R-TN) introduced the TRUMP AMERICA AI Act — a 291-page update that covered developer duties of care, how liability will work, and labor reporting requirements around AI.
The White House followed shortly after on March 20 with a four-page National Policy Framework for AI, a nonbinding but clear advice from Big Brother nonetheless. Funnily enough, on the same day, Democratic lawmakers introduced the GUARDRAILS Act to block the federal government from overriding state AI laws.
Plenty of back and forth, but there’s a lot of regulation no matter where you look. Either way, the vendors who figured out governance ahead of time are in a much better position than the ones who are waiting to see where the chips fall.
