Skip to main content
Large People ModelHuman Operating Architecture
AboutSearch the complete site3-minute readiness check

Layer 02 of 07

Decisions

Formal name: Decision Architecture

Turn choices into traceable commitments.

Understand this layer in one minute

In one minute

What decisions does

A plain-language definition, the leadership question, and the operating value of getting this layer right.

Layer job

Decision Architecture defines how choices are made, owned, documented, communicated, and revisited so the enterprise can move without losing accountability or context.

Leader question

How are decisions made and traced?

What good looks like

Important decisions have one decider, an appropriate pace, visible evidence, and a traceable commitment that survives the meeting.

01

Converts debate into commitment

A decision changes direction, timing, funding, risk, ownership, or expectations.

02

Reduces decision debt

Weak decision architecture creates repeated debate, reversal, and rework.

03

Sets AI decision boundaries

AI can support decisions, but humans must define what can be advised, automated, or approved.

Hypothetical worked example

A refund request that falls outside the standard policy

The same scenario follows readers through all seven layers. Here is the part this layer must make work.

A customer asks for an exception. A governed agent can collect the facts and recommend a response, but the organization still needs a named human owner, a decision rule, trusted information, and a reviewable record.

  1. 01
  2. 02
  3. 03
  4. 04
  5. 05
  6. 06
  7. 07

This layer's responsibility

Decisions

Decision Architecture

Operating move

A service manager decides exceptions. The agent may recommend, but it does not own the outcome.

Result across all seven layers

The customer receives a faster answer, the decision remains traceable, and AI increases capacity without inheriting authority it should not hold.

Diagnose

Start with three questions, three signals, and three measures

These are prompts for a leadership conversation, not a complete assessment instrument.

Ask

  1. 01What decisions are slowing execution?
  2. 02Who owns each decision?
  3. 03Which decisions require approval, consultation, or escalation?

Recognizable symptoms

  1. 01The same decision is debated in three recurring meetings.
  2. 02Teams cannot tell who has authority to decide.
  3. 03Important decisions live in chat, slides, or meeting memory.

Measure

  1. 01Decision latency
  2. 02Decision aging
  3. 03Decision reversal rate

Failure patterns

What weakness looks like in practice

01

Decision Fog

Everyone discusses the topic, but no one can point to the decision record.

Downstream effect: Communication increases while execution stays uncertain.

02

Authority Drift

Decision rights move informally based on personality, urgency, or hierarchy.

Downstream effect: Teams wait, escalate, or act without confidence.

03

AI Recommendation Without Rights

AI suggests actions that no human or workflow is authorized to accept.

Downstream effect: The organization either ignores AI or accepts outputs without control.

Metric signals

Measure the condition, not the activity

Decision latency

How long critical choices take to resolve.

Weak signal

Decisions stay open across several operating cycles.

Strong signal

Decision clocks and owners are visible before work begins.

Decision aging

Open decisions that exceed expected resolution windows.

Weak signal

Aged decisions create blocked work and recurring debate.

Strong signal

Aged decisions trigger escalation or scope reduction.

Decision reversal rate

How often commitments are undone or reworked.

Weak signal

Teams redo work because decision rationale was weak or missing.

Strong signal

Reversals are rare and supported by explicit new evidence.

Improve

Make three moves with one primary working tool

Keep the first intervention small enough to own, observe, and review.

  1. 01

    Classify important decisions by stakes, reversibility, and appropriate AI involvement.

  2. 02

    Assign one decider and make the roles for advice, execution, and override explicit.

  3. 03

    Record the decision, rationale, commitment, and review point in a durable place.

Primary working tool

Decision roles and rights map

Separate the authority to decide from the roles that advise, coordinate, execute, or override.

Learn how to use it

Maturity path

Progress from invisible to adaptive

Move one phase at a time. The next phase is credible only when its operating condition is observable in real work.

  1. 01

    Invisible

    Decisions happen informally and are remembered by people, not systems.

  2. 02

    Fragmented

    Some teams document decisions, but authority and evidence vary by context.

  3. 03

    Defined

    Decision types, owners, evidence needs, and revisit rules are defined.

  4. 04

    Managed

    Decision performance, aging, reversals, and escalation are measured.

  5. 05

    Adaptive

    Decision architecture adapts as strategy, risk, workflows, and AI capabilities change.

Evidence in practice

Put decisions into the work

Use concrete artifacts, bounded use cases, and visible evidence to move this layer from an idea into an operating condition.

Working artifacts

Make the layer visible

01Decision logA durable record of decisions, owners, evidence, and tradeoffs.Use when: Use for consequential choices that change direction, risk, or ownership.
02Decision rights modelA clear model for who decides, approves, consults, and is informed.Use when: Use when debates repeat or authority is ambiguous.
03Decision classification registerA register that classifies decisions by risk, reversibility, and AI role.Use when: Use before AI supports or automates decision workflows.

AI implication

What changes when AI enters this layer

What changes

AI increases decision velocity by drafting options, summarizing evidence, and recommending actions.

Primary risk

The enterprise may accept or reject AI outputs without knowing who owns the final choice.

Before scaling

Decision classes, review thresholds, evidence rules, and human override paths must be defined.

Human accountability

Humans remain accountable for consequential choices, accepted tradeoffs, and decision outcomes.

Continue with evidence

Related reading and tools

ArticleAI Agents Need Decision RightsWhy governed agents must operate inside explicit decision boundaries.

Layer connections

Read the dependency in both directions

Decisions create commitments. Weak decisions create communication overload and execution drift downstream.

Upstream condition

Ownership

Identity & Incentives

Ownership creates an upstream condition that decisions depends on.

Open layer

Current layer

Decisions

Decision Architecture

Decisions create commitments. Weak decisions create communication overload and execution drift downstream.

Downstream condition

Communication

Communication Architecture

Communication turns this layer's output into the next operating condition.

Open layer

Advanced practitioner material

Choose a topic for the task in front of you

The detailed material remains available, but it is grouped by what you need to do instead of presented as one undifferentiated list.

Start hereThe Four Decision TiersClassify decisions by autonomy, stakes, reversibility, and required human accountability before AI is introduced.
Understand the conditionBuild the minimum shared language needed to see this layer clearly. 1 topic.
Diagnose the conditionRecognize where the layer is weak, inconsistent, or creating downstream drag. 1 topic.
Design an improvementChoose structures and working practices that change how the layer operates. 3 topics.
Govern and measureMake the improvement observable, reviewable, and durable over time. 1 topic.

Apply decisions

Improve one condition in one consequential workflow.

Do not launch a broad transformation program from this page. Use the layer to make one hidden constraint visible, owned, and reviewable.

Your first working session

  1. 01

    Ask

    What decisions are slowing execution?

  2. 02

    Build

    Use the decision roles and rights map to make the condition explicit.

  3. 03

    Review

    Track decision latency and inspect the result after 30 days.

Run the diagnosticReturn to all layers