Token Unit Economics

The problem

Predicting the profitability of an AI feature is difficult when token costs, infrastructure overhead, and human support burdens are spread across different billing cycles. Calculating margins without a structured model leads to underpriced products and unexpected cloud bill shocks.

What it does

• Calculates granular token costs by separating cache-miss, output, cache-read, and cache-write buckets.
• Amortizes fixed infrastructure lumps and variable egress costs across specific unit volumes.
• Models human support costs based on per-unit minute requirements and loaded hourly rates.
• Generates a 3-axis sensitivity table covering volume fluctuations, token multipliers, and support scaling.
• Produces a standardized docs/unit-economics.md artifact for team-wide financial alignment.

Frameworks & tools

This is a methodology-driven skill compatible with any agent environment capable of file system operations (read/write) and basic arithmetic. It integrates with WORKFORCE.md specifications for multi-agent coordination.

Why this beats prompting it yourself

General prompts often conflate cache-hit rates with expensive input tokens, leading to wildly inaccurate cost projections. This skill enforces a strict procedure that prevents price invention, handles fixed vs. variable infra correctly, and includes non-negotiable legal disclaimers to ensure financial data is handled responsibly.

Use cases

• Determining the cost-per-session for a new LLM-powered chat interface.
• Evaluating how a 50% increase in token usage affects unit margins.
• Calculating the minimum volume required to cover monthly fixed hosting seats.
• Auditing the support burden of a specific feature before setting a list price.

Known limitations

Does not provide tax, legal, or investment advice. It cannot access live bank or processor accounts. It models single units only and does not perform multi-option TCO comparisons.