AI ECONOMICS · COST · ROI

Know the cost of a successful AI outcome.

Token price is not unit economics. Add retries, failures, tools, fixed software and human review — then compare the current setup with an alternative before changing the stack.

Free · no signupRuns in-browserInput values stay local

01 · ASSUMPTIONS

Current agent / workflow

Use representative monthly averages. Every assumption stays editable.

Demand & execution

Workload

Target outcomes / monthRequested completed outcomes.

Executions / outcomeBase calls before retries.

Retry rate (%)

Final failure rate (%)

AI & platform cost

Variable + fixed

Model + API / execution (USD)

Tool cost / execution (USD)

Fixed monthly cost (USD)Seats, orchestration, monitoring.

Value / successful outcome (USD)Use a conservative internal value.

Human burden

Review vs manual

Review minutes / request

Review labor / hour (USD)

Human-only minutes / success

Human-only labor / hour (USD)

Comparison configuration

Scenario B

Model + API / execution

Executions / outcome

Retry rate (%)

Failure rate (%)

Review minutes / request

HOW TO READ THE RESULT

Turn the number into a decision.

01 · COST

Start with cost per successful outcome.

This is the primary unit-economics number, not raw token price.

02 · VALUE

Check whether value survives all-in cost.

Use conservative internal value assumptions and pressure-test them.

03 · COMPARE

Compare configurations on the full system.

Retries, failures and review burden can erase a cheaper model's advantage.

04 · ACT

Escalate only when the problem is material.

High-impact workflows should also be reviewed for control, security and failure recovery.