Support Cost per Interaction: A Real Model

Most support leaders can quote a cost per agent. Far fewer can say what it costs to resolve one customer’s problem, or how that number moves once an AI chatbot takes on a share of the volume. Every AI business case, headcount plan, and vendor invoice now depends on that answer.

This eBook gives you a model to fill in, not an argument to accept. It shows you how to build a fully loaded cost per resolved issue for each channel. Then it shows you how to measure what your chatbot actually resolves, where the savings really land, and what resolution rate your AI has to reach before it lowers your costs.

This eBook also includes three worksheets, which turn each step into numbers, so you can hand them to an operations analyst and get them back filled in. The resolution benchmarks come from Comm100’s 2026 AI Live Chat Benchmark Report.

What You’ll Learn

• Why cost per resolved issue is the number to track: how the four common definitions of cost per interaction differ, and why only one of them captures the channel shifts and repeat contacts that AI changes most.
• How to measure what your chatbot really resolves: how containment, deflection, and resolution differ, and a simple way to measure resolution without re-contact using data you already have.
• Where AI savings become real cost reduction: the routes that turn freed capacity into budget finance will accept, plus the AI operating costs that have to sit in the model next to the usage fee.
• Your break-even resolution rate: a formula for the resolution rate your AI must reach to beat human cost, what each extra point of resolution is worth, and the levers that move it.

Why It Matters

AI pricing is moving at the same time that the gap between AI activity and real resolution stays wide.

• Handling is not resolving: AI chatbots handled 75.3% of chats on sites where a chatbot was deployed, but resolved 44.8% of the chats they handled (Comm100 2026 AI Live Chat Benchmark Report).
• AI resolutions are expected to cost more: Gartner predicts that by 2030, the cost per resolution for generative AI in customer service will exceed $3 (Gartner, January 2026).
• Billing definitions vary: some outcome-based pricing counts a customer going quiet after the AI’s last answer as a billable resolution.
• Finance wants proof: freed capacity only counts as a saving once a documented headcount decision is attached to it.

A cost model built on your own systems of record gives finance something to check line by line. It also gives you a margin that shows how far resolution could fall before the business case reverses.