Agentic estimating

A system proposes the calculation, a human decides whether it becomes a quote.

Definition

Agentic estimating is an estimating flow where a system proposes a calculation and its evidence, but a human confirms or rejects every step before it becomes a quote: never autonomous quoting.

What it means for the calculation

This is the guardrail that keeps a fast estimate from becoming a silent, unchecked one.

How this relates to agentic AI

This follows the broader human-in-the-loop pattern used across agentic AI systems: the system carries out the multi-step reasoning, but a person reviews and confirms consequential steps rather than the system acting fully autonomously. In estimating, that means the system can propose a calculation and its supporting evidence, but committing to a quote remains a human decision.

Common mistakes

Assuming "agentic" means autonomous. Also assuming a proposed calculation is already a binding quote before a human has reviewed it.

What it means for your calculation · most relevant in How Blake works

This is the guardrail that keeps a fast estimate from becoming a silent, unchecked one.

Read the research

Frequently asked questions

Does agentic estimating mean the system decides on its own?

No, a human confirms or rejects every step; nothing becomes a quote automatically.

What happens if a proposed calculation turns out wrong?

A human corrects it, and that correction feeds into later calibration.

Is this the same as "autonomous CAM"?

No, Subduxion explicitly excludes autonomous execution; a human always confirms.