Course contentsShow
← AI Engineering
Lesson 1854 of 1,88645. AI Product Management and IterationFree lesson

Cost per Interaction and Unit Economics

Calculate the fully-loaded cost per user interaction including inference, infrastructure, and operational overhead.

Cost per Interaction and Unit Economics

What you'll learn: You'll learn how to calculate the true all-in cost of each AI interaction to understand profitability and sustainability.


Why Cost Per Interaction Matters

Every time a user sends a message to your AI system, you incur costs: inference API calls, infrastructure compute, storage, monitoring, and human oversight. Understanding your cost per interaction is essential to determining whether your AI product is financially viable at scale.

Components of Fully-Loaded Cost

To calculate the true cost per interaction, include:

  1. Inference Costs: Token usage charges from LLM providers (input + output tokens)
  2. Infrastructure: Vector database queries, embeddings generation, caching layers
  3. Data Transfer: API bandwidth, CDN costs
  4. Operational Overhead: Logging, monitoring, error tracking services
  5. Human-in-the-Loop: Review workflows, escalations, support
  6. Storage: Conversation history, audit logs, user context

The Formula

Cost per Interaction = (Total Monthly Costs) / (Total Monthly Interactions)

For deeper insight, segment by interaction type (simple vs complex), user tier, or feature usage.

Unit Economics Example

Imagine your AI assistant costs:

  • $0.015 per inference (GPT-4 call)
  • $0.002 for embeddings lookup
  • $0.003 infrastructure + monitoring

Total: $0.020 per interaction

If your average revenue per user is 10/monthandtheyinteract200times,yourmarginis:‘10/month and they interact 200 times, your margin is: `10 - (200 × 0.020)=0.020) = 10 - 4=4 = 6 profit/user`

Key Takeaway: Fully-loaded cost per interaction reveals whether your AI product can scale profitably — track it obsessively and optimize aggressively to maintain healthy unit economics as usage grows.