We have sat in on more than forty partner pricing conversations this year. The MSPs making real margin on AI agents use one of three models — and the one you pick should depend on how your clients already think about their phone bill.
Model one: the premium seat
The simplest approach: an AI agent is a seat, priced like a seat — usually 1.5 to 2 times your human-seat price. It slots into existing quotes and invoices without explanation, which is exactly why it works. Clients already understand seats.
Best for: clients replacing an answering service or adding after-hours coverage, where the comparison is "one more employee" and the agent is obviously cheaper.
Model two: per-resolution
Charge per call the agent completes without human involvement — typically $0.75 to $2.00 depending on vertical. This aligns your invoice with delivered value and makes the ROI conversation trivial: the client can divide their answering-service bill by their call volume themselves.
The catch is billing complexity and month-to-month variance. It works best for high-volume, transactional use cases: appointment scheduling, order status, tier-one support triage.
Model three: the bundled uplift
Fold agents into a "modern communications" bundle at a flat 20–30% uplift on the whole account. No per-agent line item at all. This is the highest-margin model in practice because it reframes the conversation from cost-per-agent to outcome-per-month — and it is the stickiest, because removing the AI means downgrading the whole package.
Best for: managed-everything MSPs whose clients already buy outcomes, not licenses.
What not to do
Do not pass through per-minute AI costs with a markup. It makes your invoice unpredictable, invites line-item scrutiny, and prices your smartest product like a commodity. And do not price below your human-seat cost — it signals the agent is worth less than the humans it works beside, which undermines the sale you are actually making.