FenoriTechnologies
Selected Builds

Decision Infrastructure

A consequential decision deserves more than a score.

We built Decision Infrastructure: a system that prepares a consequential decision as a complete, evidenced argument (the strongest supportable case for and against) for a named officer to judge and sign, and retains the whole of it as a Decision Object.

StatusReference architecture
Institutional problemConsequential decisions reduced to a score
Technical foundationOpposing reasoning paths, Decision Object, model and policy versioning
Existing systems retainedExisting case and record-keeping systems
Human authorityNamed officer judges and signs by construction
Evaluated onDecision consistency, argument-evidence linkage, reconstructability
DemonstrationArchitecture walkthrough on request
DataSynthetic
DeploymentInstitution-governed AWS environment
TypeDecision systemDomainFinancial servicesStatusReference architectureComponents6

The institutional problem: consequential decisions reduced to a score.

Many systems end their work by producing a recommendation: a score with a confidence percentage. That is not enough for decisions that can affect customers, regulatory obligations or the institution itself.

A single recommendation also anchors the decision-maker around one interpretation. The uncertainty and disagreement that exist in the evidence are hidden behind the number, precisely where a regulated decision needs them exposed.

Inputs and source systems.

The system consumes a case that has already been assembled, by Case Intelligence, Shadow Case or the institution’s own workflow, and prepares the decision on top of it.

  • A prepared case and its assembled evidence
  • Applicable policy, versioned
  • The models and configurations in force, versioned

How the system is composed.

What happens when an event enters.

Prepared case + evidence
↓
Reasoning path AStrongest supportable case for
Reasoning path BStrongest supportable case against
↓
Fully argued brief
↓
Named officer reviews
↓
Human decision + signature
↓
Complete Decision Object retained

Core technical components.

  • Opposing reasoning paths
  • Evidence-linked argumentation
  • Named-signer workflow
  • Decision Object recording
  • Model and policy versioning
  • Human authority layer

The Decision Object

One reasoning path develops the strongest supportable case for the activity being legitimate; a second develops the strongest supportable case against it. Both are built from the evidence, and neither becomes the verdict. The officer receives the competing interpretations, the evidence behind each, and the points of genuine uncertainty.

Every consequential decision is stored as a self-contained record capturing the state of the decision when it was made: what was known and when, which systems were queried and what they returned, the model, policy and configuration versions in force, the evidence relied upon, the reasoning generated, both arguments, the resulting confidence, and the identity of the human signer.

Example: an exit decision, argued both ways.

An investigation into Baltika Freight OÜ concludes, and the institution faces a consequential choice: file and exit the relationship, or clear the activity. Entities and figures in this walkthrough are synthetic.

  1. The prepared case enters with its evidence: eighteen months of transaction history, the network of counterparties, two prior alerts, one prior clearance.
  2. Path A constructs the strongest supportable case for legitimacy: the trading pattern matches the customer’s declared freight business, the prior clearance examined similar activity, volumes track seasonal shipping data.
  3. Path B constructs the strongest supportable case against: the routing of three payments is inconsistent with the declared corridor, one counterparty intersects a network flagged in a related case, and the customer’s explanation for the routing has changed between interviews.
  4. The officer receives both arguments with their evidence, and the specific points where they conflict, not a score.
  5. The officer decides to file, signs, and the complete Decision Object (knowledge, sources, versions, both arguments, reasoning and signer) enters the institutional record.

How it sits inside the institution.

Evidence and audit behavior

The Decision Object is the audit behavior: a decision that cannot later be reconstructed is difficult to defend. The record survives the workflow that produced it and becomes part of the institution’s account of why it acted.

Where human authority remains

For consequential actions, the architecture requires a named officer to judge and sign. The system can prepare, reason, challenge and draft. It does not become the accountable decision-maker.

Working with what already exists

The output lands in the institution’s existing record-keeping and case systems; the signed decision travels with its full context rather than a detached score.

What the system outputs.

  • A fully argued, evidence-linked decision brief
  • A signed decision bound to a named officer
  • A complete, self-contained Decision Object retained for audit, review and challenge

What it deliberately does not do.

  • The system does not become the accountable decision-maker: for consequential actions, a named officer judges and signs by construction.
  • It does not hide uncertainty behind a single answer; disagreement between the reasoning paths is surfaced, not averaged away.
  • Preparation does not replace judgment. The officer remains the person who determines what should happen.

AI should help the institution reason. It should not quietly become the institution’s judgment.

Related builds

Built around your problem, not this one.

An institution may need this build, part of it, or something entirely different. The institution determines the outcome. The technology follows.

Discuss this outcome