Skip to content
System architecture

Built for evidence isolation, traceability, and control.

ReturnValue separates tenants, constrains access, preserves provenance, and keeps deterministic financial logic distinct from explanatory AI.

Evidence-grade intelligence · Human validation · Traceable decisions
Architecture / evidence viewIllustrative view
controlled intelligence layers
  1. EXPERIENCE PLANE · authorized investigations
  2. INTELLIGENCE PLANE · deterministic calculations
  3. EVIDENCE PLANE · isolated source context
  4. CONTROL PLANE · identity, policy, audit
  • IsolationTenant by design
  • ControlsLeast-privilege access
  • LogicDeterministic first
  • AIExplains verified facts
The ReturnValue approach

Clear boundaries create dependable intelligence.

Each layer has a defined role: receive evidence, preserve context, compute verified facts, govern investigation work, and expose only authorized results.

01

Evidence plane

Secure ingestion, validation, storage, classification, and immutable source references.

02

Intelligence plane

Versioned deterministic calculations, detection logic, confidence, and explicit limitations.

03

Experience plane

Role-aware investigations, approvals, collaboration, reporting, and audit access.

How the layers work

An architecture built for trust.

The marketing site, the application interface, the API, the data plane, and the underlying infrastructure live as separate layers. Each is independently deployable, independently observable, and bounded by an explicit contract rather than a shared object graph. The shape of the platform is deliberate: small surfaces, clear boundaries, and observability at every layer.

That separation matters when something goes wrong. A regression in an interface layer cannot escalate into a data-corruption event, because the data layer does not trust interface inputs unconditionally; it validates against the same schema that ingestion produced. A deploy to the application layer does not change the audit format, because the audit format is owned by a different service.

Workspaces are logically isolated with strong access controls and scoped data boundaries. ReturnValue is designed to use encryption in transit and encrypted storage across core systems, with platform-managed key practices. Multiple controls sit between the public surface and anything sensitive, so failure modes are contained, and backup and recovery controls are built into the architecture.

Narrow public surface
Most services are private and reachable only through a small audited gateway.
Data plane is the source of truth
No important state lives in a side service.
Versioned contracts
Internal contracts between layers are versioned, and releases stage behind contract changes so dependents do not break unexpectedly.
Observable
Key layers emit structured logs, traces, and audit events. The same trail you use is the one we use.
Connected investigation

AI cannot change the financial truth.

Generative models may explain verified evidence and assist investigation, but they do not silently manufacture calculations, records, or conclusions.

  1. 01SourcesEncrypted intake
  2. 02EvidenceIsolated storage
  3. 03LogicVerified computation
  4. 04WorkflowControlled review
  5. 05ExperienceAuthorized output
Evidence in context

Clarity at every decision point.

Architecture should make the trustworthy path the easiest path.

Illustrative view. Values shown are examples, not customer data.

Control posture

Boundaries healthy
  • Experience plane · authorized output
  • Intelligence plane · verified computation
  • Evidence plane · isolated sources
  • Control plane · identity and audit
See ReturnValue in your environment

Follow your first financial signal.

Architecture — Platform | ReturnValue.ai