According to Kaseya’s 2026 State of the MSP Report, only 13% of MSPs have turned AI and automation into a meaningful revenue stream. The opportunity is clear: clients are asking for Copilot seats, AI-driven security monitoring, and usage-based automation tools, but converting that demand into sustainable revenue requires more than adding another product to the catalog.

What fewer MSPs are talking about is what happens after the sale closes. AI products don't bill the way most MSPs are used to. A seat-based SaaS subscription is predictable: one price, one renewal date, one line on the invoice. AI consumption isn't. It's metered in tokens, inference calls, API requests — units that shift month to month, sometimes day to day, based on how a client actually uses the tool.

That mismatch is where margin quietly disappears. If an MSP's back office is still built around flat, seat-based billing, any usage that falls outside that model either goes unbilled or gets estimated — and estimates rarely favor the MSP. Estimates that err the other way are worse: an over-billed client is how an MSP loses the account. We've started calling this “AI billing leakage”: revenue an MSP has technically earned but never captures, because the infrastructure underneath the sale was never built to track it. Nobody notices a leak worth a few points of margin on any single account. Multiply that across a growing AI portfolio and a full client base, however, and it becomes a cost that never appears on the P&L but is real in what it erodes.

The instinct is to treat this as a tooling problem. You buy a metering add-on, patch the PSA, absorb it into existing billing ops. In practice, it's an infrastructure problem. Billing is only where the gap becomes visible. Before a metered AI product ever reaches an invoice, an MSP has to onboard the vendor, bundle the tool into a service it can price on outcomes, and then meter what the client actually consumed. This is the Inference-to-Invoiced journey that connects vendor onboarding, service packaging, usage measurement, pricing logic, and billing as one commercial process. Most MSP stacks do each of those steps by hand, and leakage is what that manual work looks like on the P&L. Solving it means connecting usage data—wherever it originates, across whichever vendors an MSP resells—to the catalog, the bundle and the bill, so consumption is reflected accurately instead of approximated after the fact. That's the direction the market is heading, and it's the work CloudBlue is building toward: neutral commerce infrastructure that sits underneath an MSP's stack rather than replacing it.

Tarik Faouzi

Neutral is the operative word. An MSP's relationship with its client is the most valuable thing it owns — the brand, the pricing, the trust built over years of support. Any infrastructure layer that asks an MSP to route that relationship through someone else's marketplace, or compete for the same customer, solves the billing problem by creating a bigger one. The right approach never stands between an MSP and its customer. The MSP keeps the brand, the pricing and the margin; the infrastructure stays in the background, doing the accounting.

This is also why “AI billing leakage” isn't really about AI. It's a preview. AI happens to be the category exposing the gap first, because its pricing models moved faster than most billing stacks could follow. The underlying issue is commercial infrastructure that can't keep pace with how vendors price what they sell. And that will keep resurfacing as new consumption-based and outcome-based models reach the channel. MSPs that solve it once, at the infrastructure level, won't have to solve it again every time a vendor changes its pricing model. The MSPs who win the AI era won't be the ones with the most AI in their stack — they'll be the ones who made adding the next vendor almost free.

That's the conversation I want to have with MSPs at this year's MSP Summit, under the banner of “Channel Trust.” In the end, trust follows accuracy. A client trusts an invoice that matches what they actually used. An MSP trusts a margin number it can act on in real time, not one it reconciles weeks later. Getting AI billing right isn't a side project for MSPs chasing the next shiny product — it's the foundation that makes selling AI, or anything metered, sustainable at scale. I'll be exploring exactly this on the panel Powering the Next Generation of MSPs: AI Solutions That Enable Managed Intelligence at Scale. I’ll be joined on the panel by SilverSky CISO Thomas Neclerio and N-able director of product experience Meaghan Reinecke. Jason Rinker, director of AI business development at MSP New Charter, will moderate.

Also, CloudBlue executive director Darek Tasak and I will meet MSP leaders throughout the Summit. Meet CloudBlue at Booth 424 or book a meeting with us.

Tarik Faouzi is the general manager of AI moneitziation platform CloudBlue.