Methodology

How results are produced and governed

DrillQC keeps a clear line between deterministic engineering checks and advisory AI, and keeps the evidence that makes a result reviewable.

Deterministic QC versus advisory AI

Engineering results come from deterministic checks: the same inputs produce the same output every time. AI is used only to explain and summarize results and to help navigate them. AI never produces or changes an engineering result.

Versioned inputs and outputs

Inputs and outputs are versioned so that a result is tied to the exact data that produced it. A result can be located, reviewed, and compared to earlier runs.

Reproducibility

A check is reproducible: re-running a versioned dataset reproduces the same result. Reproducibility is what makes a result reviewable rather than a one-off.

Dataset governance

Public demonstrations use controlled synthetic datasets only. Datasets move forward through explicit review rather than being changed silently to match an expected answer.

Human approval gates

Qualified people approve datasets and decisions. Approval is a human act; software surfaces the evidence, but it does not grant approval on its own.

Evidence retention

The evidence behind a result — its inputs, checks, and warnings — is retained so the result can be reviewed later, not just at the moment it was produced.

Why unresolved calculations are withheld

When a calculation is still under independent engineering review, its output is not presented as a production result. Withholding an unresolved calculation is deliberate: it prevents an unreviewed number from being treated as engineering truth.