AI Pulse · Business Models & Winners

Customer-support AI is moving from a software subscription model towards a much clearer commercial promise: pay when the system actually resolves something.

Zendesk now charges $1.50 to $2.00 per automated resolution. Intercom’s Fin charges $0.99 only when a conversation is resolved, not every time the AI takes a turn. HighRadius removed per-seat pricing entirely when it launched its new model in February. Analyst estimates suggest outcome-based structures now account for roughly 40% of deals in this category.

For a business owner, this changes the buying question. You’re no longer just asking how many agents need access, or whether another platform will fit the budget. You need to ask what “resolved” means in the contract, how it’s measured, and what happens when a customer comes back with the same issue. A low per-resolution price can still become expensive if the AI is handling large volumes of simple requests that your existing process already manages cheaply.

The upside is easier ROI tracking. If the cost is tied to completed customer outcomes, you can compare it directly with the cost of human handling, response times, and customer satisfaction. That also puts pressure on vendors to improve the quality of their automation, not just increase usage.

This is the kind of thing we build into an AI command centre, connecting AI activity to the operating measures that matter, rather than treating usage as the result.