Natural language answers with verification

Ask a business question. Inspect the proof trail.

Adequate Data turns a plain-English question into a clear answer: what the result is, what data was used, which filters applied, and how the calculation was made.

A conversation is only useful if the evidence travels with it.

The product still supports natural language questions, but the marketable promise is verification: supported answers should show their work.

Business question
Which assets were downloaded the most by sales last month?
The top assets were the enterprise deck, security one-pager, and ROI worksheet. The source, sales-team filter, date window, row count, and calculation are attached.
Can the marketing team verify that later?
Yes for this sample data: the proof file includes the proof ID, data used, generated time, filters, and assumptions. It does not prove the source system is complete forever.

Verification beats instant answers.

Adequate Data is designed to make answers inspectable, reproducible, and auditable rather than claiming that AI output is automatically correct.

Data used

Surface the source table or file, selected fields, row counts where available, and data freshness context.

Proof trail

Keep filters, calculations, assumptions, and transformation paths close to the answer so the result can be challenged.

Proof IDs

Give the answer a reviewable ID so the proof file points back to the specific work product.

Proof export

Export the evidence needed to show how a metric explanation was produced and what remains unverified.

Honest security and trust language

Adequate Data should be evaluated against documented controls: authenticated access, encrypted connections where configured, credential encryption at rest, audit logging, and proof exports for supported workflows.

Review the proof trail