Business outcomes
What does Atlas actually change for a compliance or risk team? Five outcomes, each tied to a concrete platform capability.
1. Faster, safer onboarding
A KYB onboarding that an analyst assembles by hand from three portals becomes a single durable investigation that runs the registry, ownership, screening, and adverse-media checks in parallel and returns a structured result. The analyst reviews and decides instead of gathering.
Why it holds: the seven OSINT modules run concurrently and the result is a structured report, not a pile of tabs. See the investigation story.
2. Audit-ready decisions
Every canonical value retains its competing claims and a full mutation history. When a regulator, auditor, or credit committee asks "why did you conclude this UBO / this risk level?", the provenance is attached to the value itself — including which source won, by what survivorship rule, and what the alternatives were.
Why it holds: the read path is claims-first — synthesize_entity_attributes returns the
preferred value plus alternative_claims, each with source, trust, and updated_at.
3. Consistent, tunable risk
Risk is computed by a configurable matrix engine, not hard-coded heuristics. Compliance can change policy — thresholds, dimensions, categories — and see the portfolio impact before committing, then re-score consistently across every entity.
Why it holds: matrices are data (risk_matrix_schemas), evaluated by a deterministic scorer
with live-preview and portfolio re-scoring (ADR-018,
ADR-019).
4. Repeatable enhanced due diligence
The same subject investigated twice yields a comparable, reproducible result — because an investigation is a durable workflow, not an analyst's bespoke session. Re-runs, partial-failure tolerance, and full history make EDD auditable and re-playable.
Why it holds: Temporal persists every activity; investigations resume rather than restart, and tolerate partial module failure.
5. Fewer false positives, clearer true positives
Because data from multiple providers is resolved (matched, deduplicated, survivorship-applied) rather than stacked, screening runs against one canonical entity instead of five noisy duplicates. Disagreements that matter are surfaced as conflicts for review, not silently averaged away.
Why it holds: entity resolution uses blocking + multi-algorithm similarity with explicit merge / review / create bands, and protected fields (PEP/sanctions) can never be silently overwritten.
The throughline: defensibility
Speed and coverage are table stakes — most tools have them. The outcome that compounds is defensibility: every answer Atlas gives can be traced, reproduced, and defended. That is the business case in one word.