// Glossary

The terms, defined once.

These definitions are used unchanged across the research papers, the book, the advisory work and this site. Where a term is contested or has prior uses, that is stated rather than glossed over.

The standard

Decision replayability

What is decision replayability?

Decision replayability means being able to reconstruct not only what was executed, but what the institution authorised — and to detect the difference while action is still possible.

Replayability is distinct from three things it is often confused with. Monitoring establishes what a system did. Explainability establishes why a model produced a particular output. Audit trails establish which events occurred and when. Only replayability establishes whether the decision the institution authorised is the decision that was executed.

It makes no claim about determinism and does not require an identical model output to be reproduced. It requires sufficient evidence to reconstruct how an authorised decision became a particular real-world outcome, and where any material divergence occurred.

Developed in The Irrecoverable Institution, SSRN 6714238 · Worked example → · The governance argument → · vs audit trails →

The mechanism

Decision Integrity Chain™ (DIC™)

What is the Decision Integrity Chain™?

The Decision Integrity Chain™ maps the eight layers an institutional decision moves through on its way from purpose to execution and back.

01
Purpose
02
Strategy
03
Intent
04
Rules
05
Judgment
06
Decision
07
Outcome
08
Feedback

Control is usually lost not inside a layer but at the joins between them, where intent, authority and accountability begin to separate without anything going visibly wrong. Authority is not a separate layer: authority is assigned at the Decision node, which is why authorisation alone is not evidence of control.

Developed in The Fiduciary Gap in AI-Driven Financial Institutions, SSRN 6355419 · The eight layers in full →

The exposure

The Fiduciary Gap™

What is the Fiduciary Gap™?

The structural gap between what an autonomous system is optimised towards and what the institution is obliged to protect.

Stated formally: the Fiduciary Gap occurs when the agent's executional power exceeds its contextual encoding. An institution can widen the gap without any individual system failing — by increasing what an agent is able to do faster than it encodes what the institution is obliged to protect.

Named and defined in AI Optimizes. The Bank Pays the Price., SSRN 6199979 · vs principal–agent risk →

The failure mode

Decision drift

What is decision drift?

The gradual separation of an executed action from the decision an institution authorised, occurring across the joins of the chain without any individual component failing visibly.

Drift is not error. Every component may operate exactly as designed and every control may report green while the institution arrives somewhere it did not intend. This is why component-level assurance is necessary but not sufficient, and why supervisory attention to cascading configuration change is closer to a governance question than a technical one.

Applied in the Decision Drift Audit™ and throughout When Decisions Break

The measure

Decision Velocity

What is Decision Velocity?

The relationship between the speed at which an institution can act and the integrity of the mandate encoded into that action.

Decision Velocity = (Signal Compression × Execution Autonomy) ÷ Mandate Encoding Integrity

Signal Compression is the improvement in institutional latency between detecting an event and executing a decision. Execution Autonomy is the degree of autonomous execution authority. Mandate Encoding Integrity is the degree to which fiduciary intent is expressed directly into execution constraints. The governing question is how to increase decision velocity without increasing fiduciary risk. Speed does not correct direction; speed magnifies it.

Defined in Decision Velocity and the Governance Challenge, SSRN 6355419

The consequence

Irrecoverability

What is irrecoverability in an institution?

The condition in which an institution can no longer alter the consequences of a decision, because execution has outrun reconstruction.

In an agentic institution the failure mode that matters most is not error but irrecoverability. Errors can be corrected. Irrecoverable decisions cannot: they compound, cascade, and become the institution.

Developed in The Irrecoverable Institution, SSRN 6714238

The infrastructure

Replay-Ready Infrastructure (RRI)

What is Replay-Ready Infrastructure?

What is required to make decisions replayable across interconnected institutions rather than within a single one.

RRI combines FUSE™ for immutable decision capture, STAGE™ for deterministic replay, and the Decision Integrity Chain™ extended to a network-level feedback loop. Where a decision crosses an institutional boundary no single party holds the whole chain: each can evidence its own compliance while none can evidence the decision. If settlement becomes instantaneous while governance remains retrospective, systemic fragility becomes structural.

Developed in When the Protocol Decides, SSRN 6746520

The body of work

Decision Engineering™

What is Decision Engineering™?

Decision Engineering™ examines how institutional purpose and policy become actual human and automated decisions, where that chain breaks, and how control can be rebuilt.

The term "decision engineering" is generic and has prior and parallel uses — in decision analysis since the 1960s, and more recently in software and systems contexts. What is set out here is the institutional form of it, developed by Deepak Aggarwal across eight research papers, a book and the advisory work. The un-generic assets are the ones above: the chain, the gap, and replayability as a standard.

The institutional framework in full → · Eight papers on SSRN →

The engagement

Decision Drift Audit™

What is a Decision Drift Audit™?

A ten-day engagement in which consequential decisions are followed from purpose through outcome and feedback across the eight layers, and reported as where the chain broke.

Sixty-minute executive kickoff, ten days to a board-ready finding: eight layers mapped, three exposure points identified, one Decision Integrity Blueprint™. Fixed fee.

Engagement detail →

Start with one decision

Name an important decision made in the last ninety days that your institution could not fully replay today — the rule that applied, the data that was live, the authority that permitted it. If one comes to mind quickly, these terms are describing your institution.

Test one decision → Read the research → Read the book →