Ask your managed service provider where their AI shows up on your invoice. Not in the QBR deck. On the invoice.
For most organisations the answer is nowhere, and that’s not because the provider isn’t using AI. It’s because of how MSP pricing works.
How is MSP pricing usually structured?
Strip away the packaging and most managed service contracts come down to a handful of models:
- Time and materials. You pay for the hours worked.
- A retained team. You pay for a number of engineers, whether they’re busy or not.
- Per device or per user. A flat fee per server, laptop or seat, sized on how much effort the provider expects each one to take.
- Tiered bundles. A block of hours plus response-time SLAs, with overage when you go past it.
They look different on paper, but they share one thing: they’re priced on human effort. The provider works out how many people it takes to run your estate, adds margin, and spreads it across whatever unit the contract uses. When the work grows, the people grow, and so does the bill.
That’s what we mean when we say you’re buying hours and hoping for outcomes. The provider is paid either way, whether your backups actually restore or not.
Why don’t AI savings reach your invoice?
Here’s the uncomfortable bit. Your provider can roll out AI assistants to every engineer, close tickets faster, and cut its own cost per hour, and your contract won’t change at all. You’re paying for hours, and the hour still costs what the contract says.
So the AI makes the provider more profitable, not you cheaper. That isn’t bad faith. It’s what hourly pricing does. If anything, deep automation is a threat to a business that sells hours, which is why so much “AI in managed services” turns out to be faster ticket acknowledgement rather than fewer tickets.
What does per-unit pricing change?
Per-unit pricing flips the incentive. Each host and each database carries one price. Operations hours, cost-management hours and reporting fold into the service and stop being separate invoices. The provider is now paid for keeping the estate running well, not for the time it takes.
Then you add one more rule: the unit price steps down by contract as automation rises. We measure the share of work our IT Operations Engine closes without a person and report it monthly. When it crosses the next band, the lower price applies on the next invoice. No renegotiation, no waiting for renewal.
That’s how Firemind contracts a run cost more than 30% below a traditional provider as a floor, not a hope, with the price falling further as the engine takes on more. The gain lands on your invoice, not in our margin.
How do you compare MSP prices like for like?
Comparing a per-unit quote with an hourly one is where most evaluations go wrong. A few rules keep it honest:
- Same estate, same scope. Count the hosts, databases, accounts or subscriptions, and landing zones on both sides. Include governance and reporting, tooling and licences, and FinOps and cloud-operations hours, because those are often bundled into one quote and billed separately in the other.
- Run cost only. Keep the cloud bill as its own line. Savings on the bill are real, but they shouldn’t hide a higher run cost, or the other way round.
- Look three years out. An hourly contract is flat at best. A per-unit contract with automation bands should get cheaper every year. Model both over the full term.
- Price the onboarding honestly. A one-off onboarding fee is normal. Spread it over year one so it’s visible rather than buried.
Three questions to ask your current provider
If you’d rather keep your current MSP, fair enough. Ask them for three things:
- Incidents resolved before anyone was paged, not faster ticket acknowledgement.
- Savings evidenced line by line on your bill, not a FinOps report you have to commission.
- Audit evidence produced on demand, not an annual scramble.
If you get all three, keep your provider. We mean that. If the answer is a roadmap, the problem probably isn’t their AI. It’s the model that sells you hours.
Where to go from here
The quickest way to get a like-for-like comparison is to run one on your own estate. Our pilot takes one business unit with read-only access and, in week one, maps 100% of it, finds savings of at least 3% of its bill and shows you where the dashboards are wrong. Those three checks become your TCO business case.