Cost per resolution

What is Cost per Resolution?

Cost per resolution is the total LLM spend required to resolve one user issue, used as a unit-economics metric in customer support AI. It helps teams understand what they actually pay to fully solve a request, not just to generate a response.

In practice, cost per resolution is useful because support automation often involves more than one model call, plus routing, retrieval, retries, and occasional escalation. Support vendors commonly define resolution as an issue that is fully handled without further human follow-up, which makes the metric a practical way to connect AI spend to business outcomes. (webflow2.decagon.ai)

Understanding Cost per Resolution

Cost per resolution is a unit-economics lens for support workflows. Instead of measuring the cost of a single prompt or a single conversation turn, it rolls up everything required to close the loop on one customer issue. That usually includes model tokens, tool calls, retrieval, orchestration overhead, and any fallback steps needed before the case is considered resolved.

This metric matters most when teams are trying to compare AI-assisted support against human-only workflows, or different AI architectures against one another. A low per-call token price does not always mean a low per-resolution cost, especially if the system needs multiple turns, repeated context loading, or escalations. Teams that use resolution-based billing and reporting typically care about the outcome at the issue level, not just usage at the API level. (lorikeetcx.ai)

Key aspects of cost per resolution include:

1. Outcome-based accounting: The unit is one successfully resolved issue, which ties AI spend to customer support results.
2. End-to-end cost capture: It can include tokens, retrieval, retries, tool calls, and escalation handling.
3. Workflow sensitivity: Two systems with similar model prices can have very different costs if one needs more steps to resolve the same issue.
4. Unit economics visibility: It helps teams compare support automation cost against revenue, retention, or human support cost.
5. Optimization target: It gives teams a concrete number to improve through better prompts, routing, and evaluation.

Advantages of Cost per Resolution

1. Business alignment: It connects AI spending to the result that matters most, solving customer issues.
2. Better budgeting: Teams can forecast support spend using resolved volume instead of raw request volume.
3. Architecture comparison: It makes it easier to compare model choices, retrieval strategies, and escalation policies.
4. Clearer optimization: It highlights whether costs are being driven by prompt length, retries, or poor routing.
5. Operational accountability: It gives product, support, and finance a shared metric.

Challenges in Cost per Resolution

1. Definition drift: Teams may define resolution differently, which makes comparisons hard.
2. Hidden overhead: Shared infrastructure and human review can be easy to miss in the calculation.
3. Mixed workflows: Some issues are partially automated and partially human handled, which complicates attribution.
4. Volume effects: Rare or complex issues can distort the average if the sample is small.
5. Quality tradeoffs: A lower cost per resolution is only useful if resolution quality stays high.

Example of Cost per Resolution in Action

Scenario: A support team deploys an AI agent to handle billing questions, order status requests, and password resets. Each issue may require one retrieval step, one model response, and sometimes a fallback to a human agent.

If the team spends $2,000 on model usage, retrieval infrastructure, and orchestration in a month and closes 1,000 issues with AI, the cost per resolution is $2.00. If prompt changes and better routing reduce retries and escalations while keeping the same resolution rate, that number can drop even if raw traffic stays flat.

That is why cost per resolution is useful for iteration. It shows whether improvements in prompts, knowledge access, and agent flow are actually reducing the amount of LLM spend needed to solve each issue.

How PromptLayer Helps with Cost per Resolution

PromptLayer helps teams track the prompt and workflow changes that affect cost per resolution. By versioning prompts, reviewing traces, and comparing runs, the PromptLayer team makes it easier to see which changes reduce retries, shorten conversations, and improve resolution efficiency without losing quality.

Ready to try it yourself? Sign up for PromptLayer and start managing your prompts in minutes.