Here is an awkward maths problem sitting in every managed services business right now. You bill per ticket or per hour. You just deployed AI that resolves around 65% of tier-one issues without a human. Congratulations, you have automated away your own revenue.
This is not a hypothetical. It is the central commercial question in IT services this year, and the providers who answer it early will take share from the ones who wait.
Why the Old Models Break
| Model | How it breaks under AI |
|---|---|
| Per hour | Efficiency directly reduces revenue |
| Per ticket | You are paid for volume you now prevent |
| Per device | Survives, but ignores where value moved |
| Per user, flat | Survives, margin improves quietly until a client notices |
Notice the pattern. The models tied to effort punish you for getting better. The models tied to scope do not, which is why per-user pricing is quietly winning.
The Four Models Gaining Ground
1. Outcome Pricing
You charge for a result, such as a guaranteed resolution time or an uptime level, rather than for the labour behind it. It aligns beautifully with automation and it is genuinely hard to price until you have a year of your own data. Start with one narrow outcome, not your whole catalogue.
2. Tiered Per-User With an Automation Dividend
Keep per-user pricing, then share the savings explicitly. Publish what automation removed and give part of it back as a credit or expanded scope. Clients who see the number stop suspecting they are being quietly overcharged, and suspicion is what kills renewals.
3. Platform Plus Consumption
A base fee for the managed platform, plus consumption for AI-heavy work. It mirrors how your own costs behave, which makes it honest, though clients hate unpredictable bills. Cap it.
4. Advisory and Governance Retainers
The fastest growing line for many providers. Clients now need help with AI governance, shadow AI discovery, agent security review and vendor selection. This is high-margin work that automation does not erode, because the deliverable is judgement.
The Conversation to Have With Clients
Do not hide the automation. Clients read the same vendor marketing you do, and they will eventually ask why they are paying for a hundred hours when your AI resolves two thirds of tickets.
Get ahead of it. Show the deflection numbers, show what you reinvested in proactive work, and reprice openly. A provider who raises this first looks like a partner. One who gets caught looks like a vendor.
Conclusion
Move away from any model where your efficiency reduces your revenue, and do it before a client forces the conversation. Per-user pricing with a visible automation dividend is the easiest first step, and a governance retainer is the most defensible new line. The technical work of adopting AI is the easy half. The commercial rebuild is the half that decides who is still here in three years.
Frequently Asked Questions
Will AI shrink the MSP market?
It shrinks low-tier ticket labour and grows advisory, security and integration work. The revenue moves rather than disappearing, but it moves toward providers who can do the harder work.
How do we price agent deployment for clients?
Project fee for the build, then per-user for the run. Avoid per-execution pricing early, because neither side can forecast volume and both will feel cheated.
What if clients ask for a discount because of AI?
Bring the number yourself and attach it to expanded scope. A discount conversation you initiate is a renewal conversation. One the client initiates is a procurement exercise.