AI powered legacy report reviews
Multiple clients
Technology is moving fast. Large language models (LLMs) are improving rapidly and now enable efficient first-pass summarisation, discrete data extraction, and agentic workflows. This case study shows how we are helping our clients gain better insights from historical reports faster and more efficiently, allowing them to identify and prioritise the most critical reports and information.
Several mid-tier explorers had drilling data spread across spreadsheets, PDFs, and legacy database exports. The goal was to create a single, queryable dataset that could support QA, mapping, and reporting without manual rework.
We designed repeatable input pipelines that normalised collar, survey, and assay tables, enforced validation rules, and displayed potential issues in a review dashboard. Each release produced a clean, versioned snapshot for geologists to review and approve.
Client
Mid-tier explorer
Region
Pilbara, WA
Timeline
8 weeks
Data volume
1.2M records
Challenge
- Inconsistent collar and survey formats across vendors
- Duplicate intervals and overlapping assays
- No single source of truth for QA approvals
Approach
- Built schema-mapped loaders with automated field validation
- Created QA rules and exception reports for geologists
- Delivered a versioned export and API for downstream tools
Outcome
- Reduced data loading time from days to hours
- Improved assay completeness and removed duplicates
- Enabled fast map updates and consistent reporting
Services Delivered
Tooling
Impact Metrics
Load time
3 hours (was 2 days)
Duplicate rate
0.6% (was 8%)
QA turnaround
2 days (was 2 weeks)
Teams onboarded
15 users