AI Docs Assistant Cost per User

Question

How do I estimate docs assistant cost per active user?

Quick answer

Formula: cost_per_user_month = (requests_per_user_month * variable_cost_per_request) + refresh_share_per_user_month

• Assumption: variable cost per request already includes generation, retrieval, reranking, vector, cache, and infra.
• Assumption: docs refresh is treated as a monthly fixed term and amortized across active users.
• Assumption: retrieval depth and chunking come from production-like queries, not one clean demo path.

Example: 60 requests per user at $0.031 plus $0.28 monthly refresh share gives $2.14 per active user-month.

What Usually Moves Docs Assistant Cost

• Requests per active user each month.
• Retrieved chunk count and tokens per chunk.
• Rerank depth on harder questions.
• How much of the docs corpus is re-embedded every month.

Worked Docs Corpus Example

Suppose 800 active users each run 60 requests per month. If the blended request cost is $0.031, variable usage spend is 60 * 0.031 = $1.86 per active user-month.

If monthly corpus refresh costs $224, the refresh share is 224 / 800 = $0.28. Total cost lands at $2.14 per active user-month. That is the number to compare against price, not the $0.031 request-only figure.

Where Docs Assistants Get Mispriced

• Teams count only the first answer turn and ignore clarification loops.
• Chunk count grows to rescue recall, but the new retrieval token load never gets re-priced.
• Monthly docs refresh is tracked as ops work instead of part of the product unit cost.

Related reads

• How To Choose Chunk Size and Chunk Count
• What Cache Hit Rate Means for RAG
• RAG Cost Components Explained