Framework
NET compares the value of a task's output with everything it costs to produce it, at a given level of the Task Scale.
| Part | What it captures |
|---|---|
| Value | Quality, volume and the price of each unit of output. |
| Expected loss | The cost of each way the task can fail, times how likely it is. |
| Human cost | The people time the task takes. |
| AI cost | What the AI costs to run. |
| Orchestration cost | The cost of connecting people, AI and systems. |
| Supervision tax | Overseeing, auditing and recovering from AI work. |
An AI business case that counts only the AI's running cost against saved salary is missing the supervision tax: the time people spend checking, correcting and recovering from AI output. At L5, someone approves every output. That time is real, and NET prices it.
NET is computed for a task at its current level and at a proposed level. The difference feeds the next framework, ΔNPV.