I spent last week watching RSAC coverage from every angle. Reading the keynotes, scanning the product launches, parsing the partner breakout sessions. And I kept having the same reaction I’ve had at every major industry inflection point since I started in this business two decades ago.

Everyone’s right. And everyone’s missing the point.

The Pitch

CrowdStrike CEO George Kurtz told CRN that partner organizations who figure out how to harness agentic AI “are going to do really, really well.” SentinelOne’s Tomer Weingarten called the channel “incredibly important” to scaling security operations in a world where AI agents are, functionally, new employees that need supervision. Palo Alto Networks’ Lee Klarich described a “huge opportunity” for partners to become “the trusted security architects” helping customers navigate complex AI agent deployments. Saviynt CEO Sachin Nayyar predicted that within six to 12 months, half of all companies globally will be looking for non-human identity solutions for agentic AI.

That’s CrowdStrike, SentinelOne, Palo Alto, and Saviynt all saying the same thing in the same week: agentic AI security is the next big channel revenue stream. The opportunity is real. I believe them.

Huntress co-founder Kyle Hanslovan added the punchline that stuck with me: “It’s wild how many people are missing the channel aspect.” He’s right. Smaller organizations getting hit by AI-augmented attacks can’t defend themselves alone. 1Password CEO David Faugno went further, noting that outside the tech bubble, many small businesses “don’t understand” AI and frankly resent it. Partners have to lead by example before they can lead their customers.

So the opportunity is there. The demand signal is real. The vendors are building. What could go wrong?

The Gap Nobody Talked About

SonicWall released its 2026 Cyber Protect Report on the same day I’m writing this. The timing is almost poetic. While RSAC was selling the future, SonicWall was auditing the present. Their findings are sobering.

High- and medium-severity attacks surged 20.8% to 13.15 billion hits last year. But the attacks that are actually getting through? They’re not sophisticated. Eighty-five percent of actionable security alerts came from credential and identity compromise. Not zero-days. Not AI-generated polymorphic malware. Stolen passwords.

The report identifies what SonicWall calls the “Seven Deadly Sins of Cybersecurity.” Stolen credentials as the top vector. False confidence among SMBs (88% of SMB breaches involved ransomware, more than double the rate at large enterprises). Overexposed VPN access with 82.5% growth in VPN CVEs. An average of 181 days before breaches get detected. And 44% of security alerts going completely uninvestigated due to noise.

Michael Crean, SVP and GM of managed security at SonicWall, said what I’ve been telling MSP owners for years: “The vast majority of the attacks that we’re seeing and investigating are basic fundamentals that continue to be missed.”

I’ve been in rooms where this exact conversation happens. An MSP owner comes back from a conference fired up about AI-powered threat detection. New vendor relationship. New SKU. New pitch deck. And then you look at their customer base and half of them are running VPN credentials that haven’t been rotated in 18 months, with MFA deployed on “most” accounts, which means not the ones that matter.

I’ve Seen This Before

Not the AI part. The pattern.

Every major technology shift produces a gold rush and a gap. Cloud did it. Managed services did it. SASE is still doing it. The gold rush is the new capability. The gap is the basic execution that the excitement distracts from.

When cloud took off, everyone wanted to sell cloud migration services. Great opportunity. Real money. But the partners who built durable businesses weren’t just migrating workloads. They were cleaning up Active Directory, fixing identity governance, and making sure backup and DR actually worked in the new model. The boring stuff. The stuff that prevented the 2 AM phone call six months after the migration.

Agentic AI security will follow the same pattern. The partners who win won’t be the ones who bolt on a non-human identity product and add “AI security” to their website. They’ll be the ones who build on a foundation that’s already solid. You can’t secure AI agents in an environment where your perimeter is already gone and nobody noticed. You can’t monitor agent behavior when 44% of your existing alerts go uninvestigated.

Chris Schueler, CEO of Cyderes, told CRN that non-human identity is already their fastest-growing security component. That’s a signal worth paying attention to. But Schueler also asked the right question: “How do we actually build the right control framework for agents to act, and how do we monitor their actions?” Left unchecked, agents can start acting in ways nobody authorized.

That question only has a good answer if your foundational security is tight. Identity governance. Credential hygiene. Access controls that actually get enforced. The basics that SonicWall’s data says most organizations are still getting wrong.

The Play

Here’s what I’d tell any MSP owner who just came back from RSAC:

Start two workstreams, not one. The first is your agentic AI security capability. Learn non-human identity. Understand how AI agents authenticate, what they can access, and how to monitor their behavior. This is a real market that’s forming right now, and being early matters. CrowdStrike, SentinelOne, Saviynt, and Palo Alto are all building channel motions around it. Pick one and get deep.

The second workstream is the unsexy one. Audit your existing customer base against SonicWall’s seven sins. How many of your customers have unrotated credentials? How many are running exposed VPN interfaces? How many have MFA gaps? How many have no incident response plan? Be honest with yourself. Then go fix it.

The partners who try to sell agentic AI security while their customers’ fundamentals are broken will look like the contractor who wants to install a smart home system in a house with knob-and-tube wiring. Technically possible. Practically irresponsible.

The partners who clean up the foundation first, then layer AI security on top, will own the relationship for the next decade. Because they’ll be the ones who can look a CFO in the eye and say: “Your basics are solid. Your AI deployment is governed. Here’s how we know.”

That’s not a pitch. That’s a partnership. And it’s where the real money is.