Ask questions in natural language
"What's the average grade of the Measured category above 200 mRL?" — the AI Assistant answers with the exact block-model slice, the underlying compositing rules, and the SQL-equivalent query it ran.
AI Assistant
A calibrated co-pilot for practising geologists.
Ask questions, auto-fill workflow parameters, catch anomalies and draft narratives — always grounded in your project data, always with a human in the loop.
Auto-fill workflow parameters
Get sensible starting values for iso-value cutoffs, search neighbourhoods, slope zones and revenue factors — calibrated against your project's own statistics, never hard-coded defaults.
Anomaly detection
Flag likely QAQC issues, negative kriging weights, orphan blocks, and out-of-envelope grade shells before they contaminate downstream stages.
Draft technical narratives
Generate first-draft paragraphs for your code-compliant report sections — always editable, always attributed to the human competent person for sign-off.
What a conversation looks like
You
Suggest an iso-value cutoff for the East Zone gold grade shell.
Field statistics on East Zone assays give a 55th-percentile grade of 0.42 g/t. Recommend iso-value = 0.4 g/t — this preserves 84% of contained metal while excluding barren corridors. Accept?
You
Anything unusual in the last assay batch?
Two batches (BX-2025-11 · BX-2025-14) show blank standards elevated by 3.1σ against the historical mean. Recommend re-assay or exclude before compositing. Full QAQC pack attached.
You
Draft the resource classification paragraph in a code-compliant style.
Draft prepared — 246 words. Cites slope-of-regression = 0.72 for Measured, 0.55 for Indicated. Awaiting CP review before it enters the report.
Four principles we don't compromise on
Human-in-the-loop by design
The AI Assistant proposes; the competent person disposes. Every material recommendation requires a human accept/reject before it lands in the project record.
Grounded in your data
The Assistant only reasons over your project's own statistics, tables and metadata — no hallucinated numbers, no drift into non-project trivia.
Never in customer-facing reports
Report output is 100% deterministic and free of any "AI-generated" language. External deliverables read as traditional professional engineering documents.
Full audit trail
Every Assistant interaction is logged with the exact prompt, response, dataset SHA-256 hash and the CP who accepted or rejected it. Nothing happens off-the-record.
See the Assistant on your own project.
Book a 30-minute demo — bring a dataset (or use ours) and see the Assistant reason over real drillholes.
Request Demo