The Per-Seat Model Is Dying: How MSPs Should Price in an AI-Agent World
November 1st, 2026 by Brian Suerth
Per-Seat Pricing Was a Proxy, and the Proxy Just Broke
Per-seat pricing made sense for two decades because a seat was a reasonable stand-in for work. Every employee generated tickets, needed an account, consumed a license, and created a support obligation. When a client added headcount, your revenue grew. When headcount fell, revenue fell with it - and so did the workload behind it.
AI agents break that link in both directions. One technician supervising automated remediation can cover far more users than the same technician did five years ago. Clients are also deploying agents of their own: assistants that open tickets, request access, reconcile data, and act inside systems without a person sitting at a desk. Meanwhile, a client that automates away twenty jobs may reduce its seat count by twenty while the complexity of the environment - and the support surface - goes up, not down.
The result is a pricing model that moves in the wrong direction relative to the work. Seat counts flatten or fall while the provider delivers more value, not less. Any MSP still treating users as the only billing unit is slowly handing margin back to clients in exchange for nothing.
Why the Seat Count No Longer Tracks Cost
Three things changed at once.
- Work no longer scales with people. Automated remediation, self-service, and AI-assisted triage mean a smaller team supports the same endpoints. The seats are still there, but the labor behind them is not proportional to the count
- Agents are users without seats. A client can run a fleet of automation agents against its systems with no corresponding increase in licensed users. Those agents generate incidents, need identity, need permissions, and need monitoring - real cost with no line on the invoice
- Clients see the efficiency before you reprice it. When a client cuts headcount through automation, the first invoice after the project often drops. The provider delivered the savings and then absorbed the revenue loss on top of it
What Still Correlates With Cost
Price has to attach to something that predicts effort and risk. Three measures survive the shift:
- Coverage. How many devices, sites, systems, identities, and now agents you are responsible for protecting. Coverage is durable because environments keep growing even when headcount does not
- Consumption. Tickets, incidents, automation runs, data volume, and after-hours activity. Consumption tracks effort directly, but it is volatile for the client unless bounded
- Risk. Criticality, compliance obligations, response commitments, and the cost of failure. Two clients with the same seat count can carry very different exposure
A pricing model should follow at least one of those, and the healthiest models blend two. A pure count of humans follows none of them anymore.
Four Models That Survive the Transition
Platform plus coverage tiers. Replace the per-user rate with a base platform fee and tiers based on devices, sites, and systems under management. This is the simplest migration because most clients already understand it, and it stays stable as headcount moves. The weakness is that device counts can also consolidate, so pair it with a service catalogue that defines what coverage includes.
Consumption pricing. Charge per ticket, per incident, or per automation run, often with an included allowance. This aligns cost and price most tightly, but it makes budgeting harder for the client and can punish them for having a bad month. Use it as the variable layer on top of a predictable base, not as the whole model.
Outcome-based pricing. Price against uptime, resolution time, risk reduction, or a defined business result. This is where the market is heading, but it only works if both sides can measure the outcome. Providers without reliable reporting end up arguing about the scorecard instead of getting paid for it.
The hybrid model most MSPs should actually run. A base platform fee for the stack and coverage, a device or site count that reflects the environment, a consumption allowance for tickets and automation, and a separate advisory retainer for governance, roadmap, and AI oversight. Each layer has a clear trigger, and no layer depends on a client keeping people employed.
The Efficiency Trap
AI does not improve your margin automatically. If you automate triage and keep the same per-seat price, the client keeps the savings and you keep the same revenue against a lower cost base - which looks fine until a competitor prices the same scope lower. The efficiency becomes table stakes instead of margin.
The alternative is to decide deliberately what you do with the gain. Either price the higher coverage the automation enables, sell the advisory work that makes the automation safe, or trade a lower price for a longer term and a wider scope. What you should not do is let the savings leak out silently and then discover at renewal that the client expects the lower number as the new baseline.
How to Move Without Blowing Up Renewals
- Measure cost to serve first. Build a real number per client and per service line. You cannot defend a new model if you cannot show what the old one costs you
- Quote the same scope both ways. Run a coverage-based quote alongside the current seat-based invoice. Present it at a business review, not as a surprise at renewal
- Pilot with two or three accounts. Pick clients with a good relationship and a complex environment, not necessarily the largest. Learn where the model strains before you scale it
- Grandfather existing contracts to renewal. New clients go on the new model immediately. Existing clients move at their next renewal, with the new quote delivered well in advance
- Write the scope down. Define what counts as a device, what counts as an agent, what is included in the allowance, and what triggers an overage. Ambiguity is what gets discounted later
- Train the account managers. The pricing model only survives if the person in the renewal conversation can explain why coverage and risk cost money, not minutes
The Client Conversation You Need to Be Ready For
The question is coming: if AI makes you more efficient, why am I paying the same? The answer is that the price was never for minutes. It is for the outcome and the risk you carry - response commitments, protection, continuity, and accountability when something breaks at two in the morning.
Then offer a real choice. Same price with more coverage. Lower price with a narrower scope. A longer term in exchange for a held rate. What you must not do is lower the price and keep the scope, because that teaches the client that the old number was padded.
What to Sell Alongside the New Model
Agent adoption creates a set of problems clients cannot solve themselves, and those problems are billable. Inventory and monitoring of agents, identity and permission scoping for non-human accounts, data handling rules, acceptable use policy, and cost control all need an owner. Package them as a defined advisory offer rather than answering the same questions for free in hallway conversations. The providers who do will find that the AI transition adds a revenue line instead of eroding one.
Where TAG Fits
Pricing model transitions are easier with a comparison set. TAG members benchmark their margins, service mix, and pricing approaches against peers running similar businesses, which turns a difficult repricing decision into an informed one. If you want help pressure-testing a new model, explore TAG's MSP business consulting or see what membership includes. Questions about fit or timing? Contact us - the best time to reprice is before the renewal, not after it.
Posted in: MSP Business Strategy