The execution gap in numbers
The headline finding from Omdia's research, commissioned by SuperOps, is a gap between talking about agentic AI and operating with it: 70% of MSPs report using agentic AI internally, but only 10% have extended it into real IT workflows like the service desk or security and compliance operations. Another 30% report no use at all. ITSM.tools, which analyzed the report in February 2026, describes most MSPs as being in a holding pattern, experimenting at the edges rather than deploying agents where the labor sits. Omdia estimates roughly 40% of MSP costs go to labor, which is exactly why the service desk is where the value is.
The savings case is not subtle. Omdia's modelling, as summarized by SuperOps and ITSM.tools, puts the potential at up to $1,200 per employee per year for enterprise agentic AI in employee support. For an MSP, the same logic applies to its own technicians and, more importantly, to the end-client seats it supports. The question is why a 40%-labor business is not moving faster.
An agent is only as good as the context it can see, and most MSP stacks were never built to share context.Governance is the stated barrier
Ask MSPs what is stopping them and the most common answer is governance. In the Omdia data, 47% cite governance and compliance as the top barrier, ahead of technical expertise gaps (16%), value realization and adoption (14%), data and security management (14%), and business integration (9%). ITSM.tools reads that as a signal that the technology is not the problem; what is missing is a framework for data access, transparency, auditability, and human override. We covered that angle in depth in the MSP AI governance gap and why MCP is where the numbers reconcile.
It is worth being clear about the source here. The ITSM.tools author notes that the Omdia report was commissioned by SuperOps, a vendor with a commercial interest in MSPs adopting AI-enabled IT management tools. The numbers are useful as a directional read on the market, not as a neutral census.
The barrier underneath: disconnected data
Look closely at what the same analysis says the starting point is. ITSM.tools reports that the Omdia report's first recommendation for MSPs is aligning RMM and PSA platforms so data flows between monitoring, ticketing, and billing, because without that unified foundation agentic AI lacks the context to make reliable decisions. In other words, the lead recommendation is an integration project, not an AI project.
This is where the 9% figure for business integration as a barrier is probably an undercount of the real problem. MSPs tend to file integration under "technical expertise" or "data management" because that is how the pain shows up. An agent that has to triage an alert needs to know which client it belongs to, what the client's SLA says, whether a ticket already exists, and what the work will cost. In a typical stack those facts live in four different systems with four different schemas, and each connection between them is a point of failure.
The scale of the underlying problem is visible in an adjacent data point. MSP Alert, reporting on the June 2026 SuperOps and Guardz partnership, cites a 2025 Heimdal and FutureSafe study finding that only 11% of North American MSPs report seamless tool integration. If nine in ten MSPs cannot say their tools work together, it is no surprise that nine in ten are not trusting agents to act across them.
What the vendors themselves are saying
The vendors pushing agentic AI at MSPs have reached the same conclusion. At the June 2026 SuperOps and Guardz announcement, SuperOps CEO Arvind Parthiban said, "Agentic AI depends on context," and argued that data spread across disconnected tools limits the value AI can create. Guardz CEO Dor Eisner added, "You can't safely adopt AI on top of data and identities you can't see or protect." The bundle they announced combines PSA, RMM, MDM, ticketing, and automation with an agentic security platform, and is positioned as an alternative to multivendor stacks.
That is a legitimate path for an MSP that is willing to standardize. But it answers the question for the functions that one bundle covers. An MSP serves many clients, each with its own mix of line-of-business apps, accounting systems, and security tools that no PSA vendor controls. Our earlier post on why the fix is not another all-in-one platform makes the case in more detail: consolidation shrinks the number of tools a technician opens, but it does not connect the tools a client already owns.
What an AI-ready integration layer looks like
Treating integration as the prerequisite for agents suggests a short checklist. First, a common data model, so that a ticket, a device, a contract, and an invoice mean the same thing to an agent regardless of which vendor's system they came from. Second, connections that stay correct as vendors change their APIs, because an agent acting on stale mappings is worse than no agent. Third, tenant-level isolation, so an agent working one client's queue cannot see another's data; the failure modes are covered in why an AI agent built once still breaks differently at every client. Fourth, an audit trail on every cross-system action, which is also the practical answer to the governance barrier: you can only govern what you can see.
Vendor agents have the same limit from the other direction: each one stops at its own tools, as we argued in agentic AI orchestration for MSPs. This is the design problem Ngentix was built around: a semantic model of the business data moving between systems, a runtime that watches each connection for drift, and a self-healing loop that re-maps a connector when an upstream API changes. Applied to the MSP stack, it gives agents the connected context vendors say they need without requiring every client to move onto one platform.
Why this matters for valuation and margin
ITSM.tools also reports that private equity investors and acquirers are beginning to factor AI maturity into MSP valuations, so delay carries financial risk as well as operational risk. At the same time, MSP Alert cites average net margins of about 8% versus about 18% for top performers (the article does not attribute that figure), and the SuperOps and Guardz announcement quotes a Tsquared Technologies founder saying "complexity scales faster than revenue." An MSP that connects its data first can deploy agents on the service desk with confidence, and can package the same connected layer as a service for clients, which is the idea behind offering integration as a managed service.