Exploration Insight Copilot
Outcome
Help geoscientists and mining engineers evaluate a prospect faster by unifying well logs, assay results, historical reports, and geological models into cited interpretations, with technical authority review before drilling or resource decisions.
| Portfolio attribute | Value |
|---|---|
| Primary persona | Geoscientist or mining engineer |
| Primary aggregate | ProspectEvaluation |
| Cosmos DB containers | 2 |
| Grounding data | logs, assays, reports, models, history |
| Development estimate | USD 60-220 monthly dev estimate |
| Runtime options | Python FastAPI and C# ASP.NET Core |
Productivity hypothesis
These measures reflect the workflow, evidence, decision boundary, and industry risk for this use case.
| Impact type | Measure | Conservative internal assumption | Customer measurement |
|---|---|---|---|
| Productivity | Analysis task cycle time | 8-15% reduction | (baseline median analysis-task minutes - assisted median minutes) / baseline median minutes * 100 |
| Quality or risk | First-pass acceptance | 5-10 percentage-point improvement | assisted percentage accepted without material rework - baseline percentage accepted without material rework |
Important: Conservative internal pilot assumptions only; not public benchmarks, commitments, or expected customer results. Validate against a comparable baseline or matched control cohort with mandatory human review.
Rationale: Logs, assays, reports, models, and exploration history support faster evidence synthesis for geoscientists and engineers.
Pilot guardrail: Keep geological interpretation, reserve implications, and investment decisions expert-owned.
Customer validation plan
- Baseline the current workflow with representative users and tasks.
- Run the accelerator on the same task mix with human review retained.
- Measure analysis task cycle time and first-pass acceptance with the documented formulas.
- Replace internal assumptions with customer-observed results before a business case.
Why Azure Cosmos DB
This AI solution needs an operational data plane, not only a vector index:
- Hybrid vector and full-text retrieval combines semantic recall with exact identifiers, codes, names, and domain language.
- Hierarchical partition keys and metadata predicates keep tenant and domain boundaries in the normal query path.
- Operational JSON, source chunks, workflow and agent state, feedback, and audit records remain together for low-latency reads and governed writes.
- Change feed reacts to source and operational updates without polling.
- Autoscale throughput, configurable consistency, and multi-region distribution support variable demand and enterprise availability.
Domain workflow

| Stage | Actor | Input → output | Approval |
|---|---|---|---|
| 1. Define the prospect and evaluation scope | Geoscientist | Licence area, target formation, and evaluation objective → ProspectEvaluation with data scope | Policy controlled |
| 2. Assemble subsurface and assay datasets | Data assembly service | Well logs, drillhole assays, and survey data → Quality-flagged dataset inventory | Policy controlled |
| 3. Mine historical reports for prior interpretations | Retrieval agent | Legacy reports and competent-person statements → Prior interpretations with reliability ratings | Policy controlled |
| 4. Synthesize a cited interpretation | Interpretation agent | Logs, assays, model versions, prior reports → Interpretation with uncertainty and cited evidence | Policy controlled |
| 5. Technical authority review | Chief geologist or competent person | Interpretation and model assumptions → Endorsed evaluation supporting a drilling or resource decision | Human required |
Technical architecture

Cosmos DB data model
| Container | Hierarchical partition key | Documents | Search paths |
|---|---|---|---|
prospectEvaluations | /tenantId, /prospectId | ProspectEvaluation, WellLogSummary, AssayResult, GeologicalModelVersion | /assumptionVectorEmbedding, /assumptionVector |
subsurfaceReportArchive | /tenantId, /reportId | HistoricalReport | /bodyVectorEmbedding, /bodyVector |
Query contract
| Query | Container | Scope | Purpose |
|---|---|---|---|
| Prospect dataset assembly | prospectEvaluations | single-partition | Read logs, assays, and model versions for the prospect in one partition query. |
| Ore-grade interval screen | prospectEvaluations | single-partition | Find assay intervals above the cut-off grade with passing QA/QC status. |
| Reliable historical interpretation | subsurfaceReportArchive | cross-partition-bounded | Knowledge-container query against subsurfaceReportArchive (bounded fan-out across the /tenantId + /reportId partitions named by the shortlist): retrieve legacy report content above a minimum reliability rating for citation, over the bounded report ids returned by vector shortlisting of the basin archive. |
| Model assumption drift | prospectEvaluations | single-partition | Compare model versions whose assumptions changed between builds. |
Python and C# services route each document type to its declared container and hierarchical partition key. ETag-based writes protect mutable aggregates, while append-only records and bounded cross-partition queries preserve scale.
Azure deployment

Clone the source repository, authenticate with Azure, and run:
Set-Location accelerators/oil-gas-mining/exploration-insight-copilot
az login
./Deploy.ps1 -Runtime python -Location eastus2
Change python to csharp to deploy the .NET implementation.
Explore the implementation
- Launch signed-in demo
- Implementation article
- Domain data model
- Workflow and API contract
- Python application
- C# application
- Domain container Bicep
- Platform Bicep
- Cost and configuration
- Source repository