Why AI Is Rewriting How MSPs Actually Get Paid

AI is changing MSP economics by breaking the assumption that revenue tracks headcount and seats. As automation cuts routine labor and vendors introduce usage-based AI billing, the winning business model shifts from selling technician hours to selling defined AI-enabled outcomes, priced by consumption rather than users covered.

The Real AI Story Isn't Demand. It's the Unit of Sale.

Most coverage of AI in the channel focuses on adoption curves and revenue timing. The more consequential shift is happening underneath that: the unit MSPs and their vendors actually charge for is changing. For decades, that unit was simple. A seat, a device, a user, billed monthly at a flat rate. AI breaks that math because the cost of serving a single seat is no longer fixed. A user who runs light email triage costs a fraction of one who runs long agentic coding sessions or heavy document generation, and vendors are done absorbing that variance.

Microsoft made the clearest public statement of this shift on its fiscal 2026 third quarter earnings call. CEO Satya Nadella told analysts that "the basic transformation of, I'll say, any per-user business of ours, whether it's productivity, coding, security, will become a per-user and usage business." That is not a minor pricing tweak. It is Microsoft telling its own shareholders that the per-seat model that funded three decades of channel economics is being layered with metered consumption, seat by seat, product line by product line.

GitHub followed through on exactly that pattern. On June 1, 2026, GitHub Copilot moved off flat monthly seat pricing and onto usage-based billing built on AI Credits tied to token consumption, with Enterprise seats still priced at 39 dollars a month but now including a fixed credit allowance rather than unlimited use. Heavier workflows, like long agentic coding sessions or repeated code review passes, draw down that allowance and can trigger overage charges once it runs out. A tool your technicians and developer clients treated as a flat cost is now a metered one, and GitHub built a pooling mechanism so unused credits from light users offset heavy ones across an organization, which is itself a tacit admission that per-seat billing no longer maps cleanly to per-seat cost.

The pattern is not limited to developer tools. Vendors across security, productivity, and identity are running the same experiment at different speeds, because the underlying economics are the same everywhere: a large language model call has a real, variable cost, and no vendor can keep absorbing that cost inside a flat per-seat number forever without eroding its own margin. What changes from one product category to the next is how fast a vendor is willing to say so out loud. Microsoft said it first because it has the scale and the earnings-call transparency requirements to say it plainly. Smaller vendors in the MSP stack are making the same change quietly, inside renewal paperwork, without a Nadella-level announcement attached to it.

How Should an MSP Actually Bill for AI Work?

Bill for the outcome the automation produces, not for the seat or the license sitting underneath it. If a vendor's own cost to deliver that seat now varies with usage, a flat retainer on top of it either overcharges light users or quietly eats your margin on heavy ones. The fix is not abandoning recurring revenue. It is defining the AI-enabled outcome you are actually selling, whether that is faster ticket resolution, automated patch compliance, or a specific service level, and pricing around delivering it consistently.

This is a business model problem before it is a billing problem. An MSP that keeps invoicing for "AI-enhanced service" without naming what the AI does forfeits the pricing power that specificity earns. Kaseya's 2026 State of the MSP Report, based on a survey of more than 1,000 providers, found that 53 percent are already using AI to automate high-volume tasks like ticketing, patching, and monitoring. That is real operational adoption. The report also found that average deal sizes are compressing hard, with the share of MSPs reporting more than 25,000 dollars a year from a single customer falling from 75 percent to 41 percent. Automation is not translating into pricing power on its own. It has to be packaged and sold as a defined line, not folded invisibly into an existing flat fee.

Where Does the Margin Actually Move?

The mechanics matter here, and they cut against a comfortable assumption a lot of MSP leaders still hold: that lower internal labor cost from automation shows up automatically as higher margin. It only shows up that way if your own pricing structure changes with it. If a technician-hour used to anchor your rate card and that hour now covers three or four times the ticket volume because AI triage handles the first pass, your rate card is now underpriced relative to the value delivered, not overpriced relative to cost. Leaders who treat this purely as a cost-reduction story are leaving the upside on the table instead of repricing to capture it.

That repricing work is exactly where GTIA's (formerly CompTIA's) State of the Channel 2026 research lands. The association found AI services are the top revenue growth category providers expect over the next two years, yet only about three in ten IT solution providers have fully embedded AI into how they run and price their business. The gap between naming AI as a growth driver and actually restructuring pricing around it is the gap that determines who captures the margin AI creates and who just absorbs the cost of running it.

Why Deal Size Compression Makes This Urgent, Not Optional

A shrinking deal size makes the pricing question harder to defer. When the typical account was worth well above 25,000 dollars a year, a few points of unpriced AI cost variance were survivable inside a large blended margin. At the deal sizes Kaseya's survey now shows, with the share of accounts above that threshold nearly halved, there is far less cushion left to quietly eat vendor cost swings. A business model built for larger, simpler flat-fee accounts does not transfer cleanly to a market of smaller, more numerous ones layered with variable AI cost underneath.

This is also why the fix has to happen at the business model level and not just inside a spreadsheet. A pricing adjustment bolted onto an unchanged service catalog will lag every time a vendor changes its own terms. The MSPs positioned to hold margin through this transition are the ones treating their price book as a living structure tied to defined, named AI outcomes, reviewed on the same cadence vendors review theirs, rather than a document updated once a year out of habit.

What This Means for Contracts You're Renewing Right Now

Every managed services agreement up for renewal in the next twelve months should get a clause reviewed: what happens when the vendor cost underneath a bundled service becomes usage-based instead of flat. Absorbing an unbounded vendor cost increase inside a flat MSP fee is a margin decision, not an oversight, and it should be made deliberately, with the contract language and pricing to match. This is also where operational discipline earns its keep. Providers that have already restructured client onboarding and provisioning around defined, repeatable playbooks are in a stronger position to isolate and price the AI-specific component of a service separately from everything else in the stack, which is exactly the kind of operational groundwork Catalyst is built to handle for onboarding and reducing AI-driven operational overhead.

None of this requires a wholesale rebuild of your business overnight. It requires naming, this quarter, which services in your stack now carry variable AI cost underneath them, and deciding deliberately how much of that variability you pass through versus absorb. MSPs that make that decision on purpose will set price. The ones who let it happen by default will spend the next renewal cycle explaining a margin they can't fully account for.

If you are trying to work out where your own stack has exposure to this shift, running your current lineup through the stack builder is a fast way to see which services carry AI cost variability you haven't priced for yet, and where a defined AI outcome could replace a vague line item on your next renewal.

The mechanics of AI pricing will keep shifting as vendors iterate, but the underlying discipline will not. Actiforge's full product catalog is built around named, priceable AI outcomes rather than bare seats, for exactly this reason.

See the full stack to see how each tool is packaged for a defined outcome instead of a raw license count.

Sources: Microsoft FY26 Q3 Earnings Call (Satya Nadella remarks, microsoft.com/investor) | GitHub Copilot usage-based billing announcement, effective June 1, 2026 | Kaseya 2026 State of the MSP Report | GTIA (formerly CompTIA) State of the Channel 2026 report.