The saying, “Attackers don’t sleep anymore” was what security pros used to say to justify long nights and always-on monitoring. Today, it’s closer to a literal description of the problem.
Increasingly, the “attacker” isn’t a person at all. It’s software. Not just any software, but AI systems that can probe, adapt and act on their own.
That is a fundamental change in how attacks are built and executed, not just another item on the threat landscape. It’s forcing an equally fundamental rethink of how MSPs must stop them.
At the center of that change is agentic AI.
From Tools to Actors
Most MSPs are already familiar with AI in cybersecurity, even if they haven’t always called it that. Email filtering, anomaly detection and behavior analytics have all leaned on machine learning for years. But agentic AI takes a step beyond pattern recognition.
Instead of simply flagging suspicious activity, agentic systems can set goals, make decisions and act in pursuit of those goals. Think less “smart filter” and more “entry-level analyst that never logs off.”
On the defensive side, that’s showing up in tools that investigate alerts, correlate signals across environments, and even initiate response actions.
Dawn Sizer
However, the implications are much harder to ignore on the offensive side. Attackers can now deploy systems that go beyond launching campaigns. They continuously refine them by adjusting phishing lures, mapping networks and probing for weaknesses without waiting for human input.
“What’s changed isn’t just volume; it’s the intent of the AI attack,” according to Dawn Sizer, CEO of 3rd Element Consulting. “We’re seeing attacks that don’t stop after the first failure. They adapt, retry and probe because they have unlimited time and no human limits.”
Agentic AI changes the economics of cybercrime. When attacks can scale without adding people, volume and sophistication both rise.
The Speed Problem
If there’s one word that defines agentic threats, it’s speed.
What used to take hours or days can now happen in minutes. Lateral movement doesn’t pause for coffee breaks or shift changes. Once an attacker gains a foothold, the window to respond shrinks dramatically.
“The more serious development is the growing utility of AI — and specifically agentic AI — in technical reconnaissance, vulnerability identification, rapid exploitation and attack chaining,” said Lawrence Cruciana, founder and president of CorpInfoTech. “For MSPs and our clients, that danger compounds quickly. Our valuable and often vulnerable remote access stack, administrative tooling, automation platforms, identity systems, scripts and delegated trust relationships create a kill chain that is like no other.”
That creates a familiar tension for MSPs. The number of alerts are climbing and the environments they manage grow more complex. But the available staff doesn’t increase.
“Compromise the MSP, and the attacker may gain a pathway to many clients at once,” Cruciana added.
Fighting Back with AI
For MSPs, cybersecurity is starting to feel more about holding ground than gaining it.
When it comes to the war against agentic AI, the defense is to meet automation with automation.
Across the security stack, MSPs are adopting tools that detect threats as well as help make sense of them. For instance:
- EDR and XDR: Modern platforms correlate activity across endpoints, identities, and networks. They surface what matters and suppress what doesn’t.
- Email Security: Grammar and spelling aren’t reliable signals anymore. Instead, detection is moving toward context: how a message is written, if it fits a user’s typical behavior, whether it feels just slightly off in ways that are hard to codify but easier for AI to spot at scale.
- Identity and Access Management: With more attackers bypassing endpoints and going straight for credentials, essential tools must identify suspicious login behavior or privilege escalation attempts. AI helps by spotting patterns that would be difficult to track manually across multiple tenants and environments.
Let’s not forget about response. Perhaps the most visible and controversial use of AI in security today is its role in automating what happens after a threat is detected. Some platforms can now enrich alerts, map out potential attack paths and recommend next steps. In certain cases, they can even take those steps automatically, isolating devices or disabling accounts based on predefined rules.
Lawrence Cruciana
That’s powerful. It’s also something most MSPs should approach with caution, Cruciana pointed out.
“AI cannot replace intentional and disciplined control over privileged tools, tenant boundaries, approval paths and logging.”
The Tools Behind the Trend
The vendor landscape reflects this shift in a big way. Nearly every major security platform now has an AI story. In many cases, it’s central to how the product is positioned.
Endpoint and XDR platforms are leaning heavily on AI to reduce noise and surface meaningful threats. Email security providers are using it to detect increasingly sophisticated phishing attempts.
On the network side, a multitude of vendors are applying similar techniques to traffic analysis, looking for subtle indicators of compromise that traditional tools might miss.
Identity protection is also evolving quickly to behavior-driven detection. Many of these tools are designed with AI assistants to help analysts investigate incidents, write reports or simply understand what a given alert actually means.
On the plus side, cybersecurity vendors are looking to integrate AI into existing products anywhere and everywhere it makes sense. Many MSPs won’t need to actively seek out new tools to defend themselves in all categories. Keeping up with rapidly evolving tools in the stack, however, may prove more challenging than they think.
Cutting Through the Noise
“AI-powered” has become one of the most overused phrases in cybersecurity marketing. It makes evaluation more complicated.
In practice, the presence of AI matters less than what it actually does.
- Does it reduce the number of alerts technicians have to touch?
- Does it help prioritize work in a way that aligns with real risk?
- Can it explain its decisions clearly enough that an MSP can stand behind them with a client?
Integration is another sticking point. Even capable tools that operate in isolation can create more work than it saves. The real gains tend to come when AI-driven insights can flow into existing workflows without friction.
“When we evaluate AI‑driven tools, we care less about what decisions the model makes and more about where the data goes and who else the model is talking to,” said Sizer. “If a vendor can’t clearly explain their data flows, they aren’t for this industry.”
Then, there’s the question of control. Automation is valuable, but only if it behaves predictably. Most MSPs are comfortable letting AI assist with investigation and recommendation. Fewer are ready to hand over full control of response actions, especially in environments where a false move could disrupt a client’s business.
A Measured Approach
For all the momentum behind AI in cybersecurity, most MSPs aren’t rushing toward full autonomy. If anything, the approach is incremental.
Start by using AI to reduce noise. Then, let it assist with investigation. Only after building confidence should it begin to play a larger role in response within carefully defined boundaries, Sizer warned.
“AI can identify and isolate threats in seconds. But not all systems or models can be confident with intent. Many times, that still requires a human. The understanding of the root cause can determine whether you can prevent it from happening again or if it’s a false positive.”
That caution isn’t resistance. It’s experience.
MSPs know that every new capability comes with trade-offs, and that reliability matters more than novelty. “Any MSP who lets automation investigate itself without human oversight is just moving risk around, not eliminating it,” Sizer noted.
What It’s Really About
It’s easy to frame this moment as a technological arms race, and in some ways, it is.
Attackers are using AI to move faster and operate at greater scale. Defenders are doing the same to keep up. But the more practical way to think about it is simpler than that.
This isn’t really about AI at all. It’s about leverage.
The number of threats isn’t falling, the complexity of client environments isn’t shrinking and the pool of available security talent isn’t suddenly expanding. When it works as advertised, AI gives MSPs a way to stretch their capabilities without a corresponding increase in workload.
In a world where the attackers are increasingly autonomous, that kind of leverage is essential.
There’s no doubt that MSPs will use AI, Cruciana concluded.
“The question is whether they will apply it inside a system that prevents their own operational leverage from becoming the attacker’s advantage.”
AI Defense Roundup
As ChannelPro’s online director and tech editor for over a decade, Matt Whitlock has spent years blending sharp tech insight with digital know-how. He brings more than 25 years’ experience working in the technology industry to his reviews, analysis, and general musings about all things gadget and gear.
Images: InfiniteFlow — stock.adobe.com, everythingpossible — stock.adobe.com
