What is evidence-to-decision infrastructure?

Subsurface resource decisions (oil and gas acquisition, mining investment, geothermal development, carbon storage review) all share a structural problem. The evidence is fragmented, incomplete, and costly to verify. The decision is consequential. The reasoning has to hold up under scrutiny.

Take an oil and gas acquisition. The economics of a bolt-on package rest on one figure: how much each well will produce. That figure arrives as a seller's type curve, built to sell, while the wells that could test it, the offsets already producing nearby, sit scattered across thousands of public filings nobody has time to assemble before the committee meets. The evidence exists. The coherence does not.

Evidence-to-decision infrastructure is the layer that connects raw evidence to the claims, reviews, approvals, and revisions that produce a defensible decision. It is not a report generator, a geological model, or an autonomous AI. It is the controlled workflow, evidence lineage, and institutional memory that makes the decision reviewable.

Three layers

The product has three distinct layers:

1. Platform. The durable evidence, provenance, review, decision, monitoring, and revision infrastructure. This does not change between domains.
2. Shared workflow. An opportunity moves from source evidence through claims, contradictions, gaps, expert judgment, approval, and later revision. This is the same shape across subsurface decisions.
3. Configured application. Domain questions, terminology, stage criteria, decision policies, and templates that make the shared workflow useful for a specific decision. Oil and gas deal screening is the first, and the first live configuration. Mining investment diligence, geothermal, and carbon storage are candidates.

The architecture is vertical at the workflow and reusable underneath. Each configuration earns its own domain rules and validation evidence.

Why not just use a language model?

If excellent general-purpose AI became nearly free tomorrow, would customers still need this infrastructure?

Yes, because access to a language model is not a moat. A model can regenerate an answer. It cannot reconstruct the trail of which source a figure came from, who reviewed it, what they disagreed on, and what has changed since. That trail has to be built while the work happens, not afterwards.

The product owns the controlled workflow, evidence lineage, institutional context, approvals, and history of changing decisions. The AI is a set of replaceable capabilities (extraction, comparison, retrieval, drafting, monitoring) attached to the workflow, not a separate authority layer.

What it is not

• Not an autonomous geologist or a replacement for qualified professional judgment.
• Not a universal geological knowledge graph or a global resource-potential map.
• Not an opaque project score that hides its reasoning.
• Not a report generator. The application state is the product; the memorandum is one rendering of it.

The product promise

Know what supports the decision, what contradicts it, and what could change it.

Evidence first. Decisions that remain reviewable.

Join the waiting list to follow the build.