Platform

A platform designed to govern recovery actions

AxiomCausal is designed for organizations that need local deployment, operational control and decision-level explainability.

Architecture

The decide() pipeline

A pure function that takes a frozen state and returns a ranked recommendation with a complete constraint trace. Five components, one audit spine, zero hidden state.

The decide() pipeline — five components, one pure function
AxiomCausal decide() architecture A horizontal pipeline of five components. From left to right: L1 Counterfactual simulation, L2 Causal effect estimation, L3 EBC module application, L4 Safety gate evaluation, L5 Message generation. Arrows connect each component to the next. A downward arrow from the pipeline leads to the Append-only audit spine. L1 Counterfactual simulation L2 Causal ITE (DML / forest) L3 EBC modules (7 constraints) L4 Safety gate (human review) L5 Message generation Append-only audit spine DB trigger enforced · full constraint trace T

decide(state, t) is a pure function. Every invocation writes a constraint trace T to the append-only audit spine, so any historical decision can be reproduced by replay.

Walk-through

A 90-second agent flow

Case selection, policy decision, explain, audit trail, safety gate, timeline. Narration is deliberate: no voice-over, captions only.

Demo video being prepared.
A 60-90 second walk-through of the agent flow (case selection, policy decision, explain, audit trail, safety gate) is being recorded. Until it is published here, the one-page overview PDF carries the equivalent architectural summary in text form.

Full-screen recommended. Ambient audio only.

Agent surfaces

What operators see

Five screens that carry the product's auditability claim end-to-end. All screenshots are captured against a running frontend with anonymized seed data.

Case Detail — Policy Decision
The top-ranked action shown alongside the policy trace that justified it. Every EBC module that contributed is named.
[PLACEHOLDER · slug=case-detail-policy-decision]
TODO: case-detail-policy-decision screenshot to be captured via Playwright against a running frontend agent with anonymized seed data. See docs/commercial/BLOC3_PLACEHOLDERS.md.
Audit Timeline — Append-Only
Every action, every override, every escalation — written to an append-only spine enforced at the database trigger level.
[PLACEHOLDER · slug=audit-timeline-append-only]
TODO: audit-timeline-append-only screenshot to be captured via Playwright against a running frontend agent with anonymized seed data. See docs/commercial/BLOC3_PLACEHOLDERS.md.
Explainability — Constraint Trace
For any past decision, the full constraint trace T is reconstructible by replay. Not a feature-importance plot.
[PLACEHOLDER · slug=explainability-constraint-trace]
TODO: explainability-constraint-trace screenshot to be captured via Playwright against a running frontend agent with anonymized seed data. See docs/commercial/BLOC3_PLACEHOLDERS.md.
Policy Module Configuration
EBC modules can be enabled or disabled individually. Each is named, each has an explicit validation status.
[PLACEHOLDER · slug=policy-constraint-configuration]
TODO: policy-constraint-configuration screenshot to be captured via Playwright against a running frontend agent with anonymized seed data. See docs/commercial/BLOC3_PLACEHOLDERS.md.
Safety Gate Dashboard
Human review triggers visible in real time. The gate is a dispatch constraint, not a recommendation layer.
[PLACEHOLDER · slug=safety-gate-dashboard]
TODO: safety-gate-dashboard screenshot to be captured via Playwright against a running frontend agent with anonymized seed data. See docs/commercial/BLOC3_PLACEHOLDERS.md.

Governed action

The platform supports action selection that remains readable in business terms and reviewable when scrutiny increases.

Policy-driven escalation

Recovery actions follow a governed escalation path defined by reviewable business rules.

Business-readable logic

Decision logic remains understandable to non-technical stakeholders.

Controlled overrides

Manual overrides are tracked, justified and auditable.

Behavior-aware communication

Communication is handled as a governed operating layer, not only as a reminder queue.

Contextual messaging

Communications adapt to debtor context and behavioral signals.

Channel governance

Channel selection follows policy, not convenience.

Quiet hours and protections

Rate limits and timing controls are built into the operating model.

Decision accountability

Performance is attributable and understandable. Accountable decision support, not opaque output.

Causal measurement

Holdout-based measurement isolates platform impact from background recovery.

Justification logs

Every decision is logged with full input context and policy trace.

Audit-ready reporting

Reports are generated with traceable data and hash-verified integrity.

Operational fit

Built to fit institutional constraints around ERP, deployment, security and reviewability.

ERP integration

Webhook-based connectors for Odoo, SAP, Sage and generic ERP systems.

Sovereign deployment

On-premise, private cloud or air-gapped deployment options.

Multi-tenancy

Tenant isolation for organizations managing multiple entities or subsidiaries.

Not another collections UI

AxiomCausal should be evaluated as an institutional decision platform for recovery operations.

Explore deployment and governance

Understand how AxiomCausal fits sovereign deployment requirements.

View Sovereignty