The Continuity Series · Companion Documents

The Delegation Framework

The Law · The Index · The Practice

Henning du Preez · 2025

I The Delegation Law
II The Delegation Readiness Index (DRI) — v1.0
III DRI in Practice — v1.0
Version 1.0 · CC BY 4.0
What This Is

Three documents. One governing rule.

Read in sequence, or enter at any point. Each document stands alone. Together they form a complete framework for deciding what may safely be delegated to machine intelligence — and what may not.

Document I
The Delegation Law

Names the single governing rule. Defines what delegation must not do. Establishes the three dimensions of readiness.

Document II
The DRI

Operationalises the law into a non-compensatory index. Each dimension must hold independently. Failure in any one fails the whole.

Document III
DRI in Practice

Four grounded examples — scheduling, strategy, defensive automation, and norm enforcement — showing the framework applied to real decisions.

Document I
The Delegation Law
The single governing rule for delegating decisions to machine intelligence
The Problem

Every major system now faces the same question.

Not whether machine intelligence is powerful.
That is already settled.

The question is whether decisions should be delegated to it.

Until recently, this question did not exist. Machines executed instructions. Humans decided. That boundary is gone.

Reasoning itself can now be performed by machines. At speed. At scale. Without fatigue.

This creates a new problem. Delegation is no longer a technical choice. It is a structural one.

History shows that systems do not fail because they adopt new power. They fail because they delegate without a governing rule.

This document names that rule.

The Law

One rule. No exceptions.

The Delegation Law

A decision may be delegated to machine intelligence if and only if doing so does not degrade the integrity of the system over time.

Nothing else matters.

Not capability.
Not efficiency.
Not competitive pressure.
If integrity degrades

Progress is temporary. The system decays at a pace that may not be visible until reversal is impossible.

If integrity holds

Progress compounds. Scale strengthens the system instead of hollowing it out.

What This Law Exists to Prevent

Delegation fails in predictable ways.

Sometimes a decision should never have been delegated. Sometimes it works at first, then collapses as reliance spreads. Sometimes it succeeds locally while corroding the system quietly.

These failures are not moral. They are structural. The law exists to surface them before delegation becomes irreversible.

How the Law Is Judged

The law is enforced by readiness — not intention.

Readiness has three independent dimensions. Strength in one does not compensate for weakness in another. If any one fails, delegation fails.

Structural Alignment

Does delegation remove a real limitation, or merely automate around it? If the underlying constraint remains, delegation only accelerates failure.

Propagation Stability

Does the decision remain sound as reliance spreads? If behaviour degrades with scale, the system is not ready.

System Integrity

When the machine is wrong, misused, or attacked, does the system recover? If errors compound silently, delegation is unsafe.

The Non-Compensatory Rule

The three dimensions are not scored and averaged. They are each independently required. A system scoring perfectly on two dimensions and failing on one is not 67% ready. It is not ready.

Scope

What the law is not.

It does not ask whether delegation is impressive. It asks whether it is survivable.

Why the Law Must Be Explicit

When a rule is implicit, delegation defaults to capability. Capability always outruns readiness.

Systems then learn too late. Corrections arrive after damage is done. Control is replaced by reaction.

An explicit law changes this. It allows restraint without weakness. Progress without denial. Leadership without heroics.

Consequence

Two futures. One structural choice.

Without the law

The system does not fail loudly. It decays. Judgment erodes. Responsibility diffuses. Recovery becomes impossible. By the time collapse is visible, delegation is already total.

With the law enforced

Delegation becomes durable. Human judgment does not disappear — it relocates. From execution to design. From reaction to supervision. From heroics to structure.

Closing

Machine intelligence makes delegation inevitable.

This law makes it governable. Progress will continue either way. The only open question is whether it will endure.

Document II
Delegation Readiness Index
DRI v1.0 · Operationalising the Delegation Law into a decidable assessment
Purpose

From law to decision.

The Delegation Readiness Index exists to operationalise The Delegation Law. It provides a clear test for determining whether a decision may be delegated to machine intelligence without degrading the integrity of the system over time.

What DRI Measures

The index does not measure capability. It measures readiness. These are not the same thing. A system may be highly capable of performing a decision and still not be ready to hold it durably.

Definition

A non-compensatory index.

DRI Formula
DRI = Structural Alignment × Propagation Stability × System Integrity
Each dimension must hold independently. The index is multiplicative by design.
Strength in one dimension cannot offset weakness in another.

Failure in any single dimension results in delegation failure. The multiplication is not mathematical — it is structural. A zero in one column makes the product zero, regardless of the others.

The Three Dimensions

Each dimension is independently required.

Dimension I
Structural Alignment
Does delegating this decision remove a real system limitation, or merely automate around it?
Holds when
  • The governing constraint is invalidated, not bypassed
  • Delegation increases system effectiveness, not just efficiency
  • Human bottlenecks are removed rather than relocated
  • The old rule no longer governs outcomes
Fails when
  • Delegation accelerates an unchanged constraint
  • Human judgment remains the hidden bottleneck
  • Errors increase faster than insight
  • Speed substitutes for understanding
If Structural Alignment fails, delegation amplifies failure.
Dimension II
Propagation Stability
Does the decision remain sound as reliance spreads across users, systems, or time?
Holds when
  • Outcomes remain stable under increased adoption
  • Misuse does not become the dominant mode
  • Behavioural reversion is limited and detectable
  • Coordination cost does not grow faster than value
Fails when
  • Local success collapses at scale
  • Dependence introduces new systemic risks
  • Human vigilance erodes as delegation becomes normal
  • Edge cases become the norm
If Propagation Stability fails, delegation destabilises the system.
Dimension III
System Integrity
When the machine is wrong, misused, or attacked, does the system recover?
Holds when
  • Errors are bounded and visible
  • Decisions are auditable and attributable
  • Recovery paths exist and are exercised
  • Dependency does not hollow out human judgment
Fails when
  • Errors compound silently
  • Accountability diffuses or disappears
  • Reversal is impractical or impossible
  • Long-term capability decays
If System Integrity fails, delegation causes irreversible damage.
Evaluation Rule

The decision rule is simple.

A decision may be delegated to machine intelligence if and only if:

If any dimension fails, delegation is structurally premature — regardless of performance or demand.

What DRI Is Not

DRI is not a benchmark. It is not a maturity model. It is not a safety checklist. It is not a prediction of outcomes. It does not rank systems. It governs decisions.

Use

Applied at the point of delegation.

DRI is applied before reliance becomes normal. It is intended for: system designers, organisational leaders, regulators, boards, and anyone accountable for long-term outcomes.

DRI does not remove responsibility. It makes responsibility explicit.

Delegation without readiness compounds risk.
Delegation with readiness compounds value.
Document III
DRI in Practice
v1.0 · Four grounded examples applying the Delegation Readiness Index to real decisions
What This Section Is

Calm, concrete, and usable.

This section shows how to apply the Delegation Readiness Index to real delegation decisions using four grounded examples. It is designed to be approachable — especially when the decision feels urgent.

How to Use DRI · In One Minute

1. Structural Alignment: Are we removing the real constraint, or just automating around it?

2. Propagation Stability: As reliance spreads, do outcomes remain sound?

3. System Integrity: If it's wrong, misused, or attacked — can the system recover cleanly?

If any dimension fails, delegation is structurally premature. Strength in one cannot cover weakness in another.

What the Examples Will (and Won't) Do

Each example is intentionally short. It does not try to convince you of a worldview. It simply shows how DRI makes the decision decidable.

Notice how often the failure mode is not "the system isn't smart enough," but "the rule is not finished" or "the recovery path is not real."

Four Examples

Why these four were chosen.

Each example pressures a different part of the framework:

Example 1
Operations Scheduling Scale effects
Works Small, Weakens the System at Scale
Delegation candidate: Using an AI system to automatically schedule customer service agents based on demand forecasts.

A small team adopts automated scheduling to reduce manager workload and keep coverage aligned to expected demand. Early results look good. Reliance grows.

Structural Alignment Holds

What it removes: manual scheduling effort and delays.

What remains: the business rules for coverage are reasonably clear and do not depend on human judgment at the decision point.

Propagation Stability Marginal

As the team grows, edge cases multiply — fatigue, fairness, skill mix, training time, personal constraints. The system optimises "coverage," but important human signals begin to be treated as noise. Over time, people stop raising issues because "the system decided."

System Integrity Fails

Bad schedules often show up late — burnout, morale decay, resignations. Accountability blurs: "Who approved this schedule?" Recovery is possible, but usually after cost has already been paid.

DRI Verdict
Not delegable at scale
The delegation removes effort, but quietly erodes responsibility and recovery. A safer path: keep the automation narrow (suggestions + human sign-off) and treat "recovery time" as a first-class metric.
Example 2
Leadership Strategy Judgment
Delegating Company Strategy
Delegation candidate: Letting an AI system define company strategy based on market data, trends, and simulations.

A leadership team wants to reduce bias and speed up decision-making. They propose "data-driven strategy" where the system selects markets, positioning, and priorities.

Structural Alignment Fails

The real limitation: uncertainty about the future — not lack of information.

The missing prerequisite: a stable rule that defines what "correct strategy" is. Strategy commits the organisation to a chosen future under uncertainty. That commitment requires a human to be responsible for it.

What strategy actually does is not computable. The system cannot hold the accountability that the commitment creates.

Propagation Stability Not evaluated

Structural Alignment failure halts delegation. The remaining dimensions are not evaluated.

System Integrity Not evaluated

Structural Alignment failure halts delegation.

DRI Verdict
Structurally premature
The safer use of machine intelligence here is support, not delegation: improve visibility, model scenarios, and surface second-order consequences — while humans remain accountable for the strategic commitment.
Example 3
Defense Reaction time Strict scope
Narrow, Rule-Bound Defensive Automation
Delegation candidate: Automatically triggering a defensive intercept when an incoming projectile meets a strictly defined threat profile.

Human reaction time is insufficient to respond safely. The system is designed to operate in a narrow, well-defined envelope with explicit constraints.

Structural Alignment Holds

What it removes: the reaction-speed limitation — a genuine structural constraint that humans cannot overcome physically.

What remains human: rule definition, scope, oversight, and accountability. The system executes a fixed rule. It does not invent one.

Propagation Stability Holds

Scaling increases consistency, not ambiguity. Misuse is constrained by the system's strict operating envelope. The rule does not drift simply because adoption increases.

System Integrity Holds (within strict bounds)

Actions are logged, attributable, and auditable. Clear disable mechanisms and safe modes exist. Scope is narrow enough that failure does not spill into unrelated domains.

DRI Verdict
Delegable — within a narrowly defined scope
This example makes a non-obvious point: delegation can be structurally safe even in serious domains, when the rule is explicit, the scope is tight, and recovery is real.
Example 4
Society Norms Rule drift
Delegating Cultural Norm Enforcement
Delegation candidate: Using AI to automatically detect and enforce "acceptable speech" or "approved behaviour" across a platform or society.

The intention is often understandable: reduce harm, reduce conflict, and keep environments usable. The challenge is that norms are not static — they evolve across communities and time.

Structural Alignment Fails

The real limitation: ongoing disagreement about values and meaning — not enforcement capacity.

The missing prerequisite: a stable rule that stays legitimate across contexts. Without it, the enforcement system effectively becomes a rule-maker — which is a different, and far more consequential, function than enforcement.

Propagation Stability Not evaluated

Structural Alignment failure halts delegation. The remaining dimensions are not evaluated.

System Integrity Not evaluated

Structural Alignment failure halts delegation.

DRI Verdict
Structurally premature
A safer approach is governance-by-visibility: make rules explicit, allow appeal, preserve accountability, and keep humans responsible for the rule — even if machines assist with detection and routing.
The Practical Lesson

Not "no forever" — "not yet."

Delegation becomes safe when the rule is finished, the scope is bounded, and recovery is real.

When those conditions are missing, the most constructive move is usually not refusal — it is clarity about what must become true to make delegation ready. The framework is not a brake. It is a map.

Reader Safety (Intentional)

This framework is written to keep the reader safe to think. It avoids urgency language and avoids framing the reader as "at risk" for asking the question. DRI is a decision aid, not a moral judgment.

Machine intelligence makes delegation inevitable.

The Delegation Law makes it governable.

DRI makes it decidable.

↑   Back to Top