MSP 1.0 was break-fix. MSP 2.0 was managed services. MSP 3.0 was per user, per device. AI has just forced MSP 4.0 — and on the panel at SMBiT recently, nobody had a clean answer for how you price it. Me included.
(MSP = managed service provider. The company that looks after your IT.)
The trap
Every model our industry has ever used charged for inputs — hours, seats, devices. An AI agent that closes a ticket before a human reads it shrinks the input to nearly zero while the value to the customer stays exactly the same, or goes up.
Price on inputs and we end up charging less for doing more. Price on outcomes and we have to get good at measuring something we’ve never had to measure.
What an outcome actually looks like
- Tickets resolved before a person touches them
- Downtime avoided, not downtime fixed
- A staff member back working in minutes, not tomorrow afternoon
The shift is coming faster this time
When we started 20 years ago we were 100% hourly. Today we’re 90% contract, priced per user. That shift took us the best part of two decades — AI is going to force the next one a lot faster than that. We’re working through what it means for our own pricing right now. Not finished. Far more useful to argue about it in a room full of people running the same business than to guess on my own.
If you run an MSP — how are you pricing AI work today? And if you buy IT services, go and look at whether your agreement pays your provider for hours or for results.
Thanks to the SMBiT team for having me on the business panel.