The AI Revenue Line Most MSPs Are Still Counting Wrong

AI services revenue is growing fast for MSPs, but most of it is not monthly recurring revenue yet. Recent industry research shows client demand for AI services far outpacing meaningful revenue capture, and much of what does convert lands as a one-time implementation fee rather than a subscription line. Count it as MRR before it behaves like MRR, and your recurring revenue number stops meaning what you think it means.

Is AI revenue actually MRR right now?

For most MSPs, no. The 2026 Kaseya State of the MSP Report, based on responses from more than 1,000 MSPs worldwide and released in April 2026, found that 48 percent of MSPs rank AI and automation as the top client need for the year, ahead of security and backup. Only 13 percent report generating meaningful revenue from those same services. That gap is the whole story. Demand is real. A recurring, billable line behind it mostly is not, at least not yet.

This matters because the way you answer that question determines whether your MRR discipline holds up under scrutiny. A board deck, a lender, or a buyer doing diligence on your book is going to ask what is driving your recurring revenue growth this year, and "AI" is an increasingly common answer. If the honest answer is that most of that AI revenue is a handful of assessment engagements and one-time Copilot rollouts, you have project revenue wearing an MRR label, and that is exactly the kind of quiet mislabeling that MRR discipline exists to catch.

Why is there such a wide gap between AI demand and AI revenue?

The gap exists because most AI work MSPs are doing right now is inherently a one-time event, not a subscription. The GTIA State of the Channel 2026 report found that in DACH and the UK & Ireland, more than a third of ITSPs report that 11 to 25 percent of their revenue is currently generated from AI-related products and services. Consulting and structured assessments are a real entry point, and they are genuinely useful for building client trust. They are also, by design, revenue you earn once per client rather than every month.

Separately, Informa's 2026 MSP 501 Survey Insights Report, sponsored by Ingram Micro, found that 57 percent of MSPs now expect AI to rank among their biggest revenue gainers over the next 12 months, up from 37 percent a year earlier, with nearly a third projecting AI revenue growth above 20 percent. That optimism is not wrong. It is just measuring something broader than MRR. Revenue gains and recurring revenue gains are not the same claim, and a survey question about expected revenue growth does not ask respondents to separate the two. If you are pulling that kind of stat into your own board narrative, you need to know which one you are actually reporting.

What changes when you separate implementation from the recurring line?

Everything about how the number behaves. Implementation work, whether that is a readiness assessment, a Copilot rollout, a custom automation build, or an initial data governance review, is priced once and delivered once. The Scopable guide to MSP AI services pricing makes this split explicit, separating one-time project work (a readiness assessment, a Copilot rollout, a governance package, and workflow automation projects) from recurring, ongoing support (quarterly reviews, policy updates, user training refreshes, and usage reporting). Only the second category behaves like managed services revenue, meaning it renews without a new sales conversation and it survives a client not calling you this month.

Here is a simple version of that split, adapted for how most MSPs are structuring AI engagements right now.

If your books are only tracking a single "AI revenue" line, you cannot see this split, and you cannot tell your board or a prospective buyer how much of that growth is durable. Break it into at least two lines, one-time and recurring, before you report on it again.

How does this affect your valuation multiple?

It affects it directly, because buyers and lenders are not pricing your revenue growth story, they are pricing your recurring revenue base. An AI practice that looks like fast growth on a topline chart but is mostly assessment fees and rollout projects will not move your MRR share, and MRR share is what gets rewarded in a deal. Reporting AI as a growth driver without disclosing that most of it is non-recurring is the kind of thing that gets caught in diligence, not before it. Better to catch it yourself first, in your own reporting, than to have someone else catch it during a term sheet review.

Building the recurring AI line MSPs can actually count

The fix is not to slow down on AI. It is to be deliberate about which parts of the offering you structure as a subscription from day one, instead of hoping a project client converts into a retainer later. That means pricing ongoing monitoring, model update management, and AI tool governance as a standing monthly fee attached to the engagement, not as a follow-on you pitch after the project wraps. It also means giving your sales team a packaged version of that recurring layer to sell alongside the one-time work, so the retainer is part of the original proposal rather than a separate ask.

It also lines up with where MSPs say they are already focused. The ScalePad 2026 MSP Trends Report, based on more than 1,100 MSP professionals surveyed across North America, found that growing existing client accounts jumped from the fourth most common growth strategy last year to the second this year. Clients who already paid for an AI assessment or rollout are the easiest group to convert to a recurring AI governance retainer, since you are already in their environment and they already trust the initial work. Treat that conversion step as part of the sale, not an afterthought.

This is precisely the seam that white-labeled tooling is built to close. Tools that treat AI operational overhead as a packaged, ongoing service layer, rather than a bespoke build for every client, turn what would otherwise be a one-time project into a line you can actually forecast. Catalyst is built around exactly that idea, taking the operational load of running AI tools for clients and making it something you resell as a standing service rather than rebuilding it project by project.

If you are not sure how much of your current AI revenue is genuinely recurring versus still riding on one-time engagements, it is worth running your current offering mix through the stack builder to see where the gaps between project work and packaged recurring service actually sit. And if AI is not the only place your recurring revenue mix needs tightening, the rest of the product catalog is built around the same principle, turning things you are already doing manually into something you can sell every month instead of once.

MRR discipline was never about refusing new revenue lines. It is about being honest, in your own reporting, about which lines are actually recurring before you let them inflate the number you use to run the business. AI is the newest test of that discipline, and right now, most of the channel is still failing it quietly. See the full stack.