1
Who owns outcomes?
Canonical Reference
v1.0LPM Framework Specification
The single, versioned source of truth for the Large People Model. Every other page teaches these definitions. This page fixes them.
How to use this specification
Use this page to verify an exact definition, count, sequence, or framework relationship. It is the canonical reference, not the recommended starting point. Start with the Framework when you need orientation or diagnosis.
Start with the FrameworkApproved concept hierarchy
- Four Executive Questions
- Define what leaders need the operating model to answer.
- Two Axioms
- State the non-negotiable premises for sequencing and human accountability.
- Seven Layers
- Show where to look for an operating condition or constraint.
- Nine Laws
- Explain why operating patterns repeat and interact across layers.
- Five Maturity Phases
- Sequence how the organization builds readiness over time.
Changelog
- v1.02026-07-03 Initial public specification.
Definition
What LPM is
The Large People Model is a Human Operating Architecture: a structured model of how ownership, decisions, communication, information, platforms, governance, and AI operate as one system.
It exists to make the hidden operating model visible, governable, and safe to automate, so an enterprise can absorb AI without scaling confusion.
Category: Hybrid Workforce Architecture and Control
What it is not
- Not AI governance tooling.
- Not an agent-building framework.
- Not observability or monitoring.
Foundational Truths
The two axioms
Sequencing
You cannot govern what you have not owned. You cannot automate what you have not governed.
Accountability
A tool cannot own an outcome. A human always does.
Structure
The seven layers
A dependency chain, not a checklist. Each layer shapes the one above it.
01Identity & IncentivesDefines who people are in the operating model, what they own, and what behaviors the organization rewards.02Decision ArchitectureDefines how decisions are made, owned, documented, communicated, and revisited.03Communication ArchitectureDefines how context, priorities, risks, updates, and decisions move across the organization.04Information EcologyDefines how data, knowledge, documents, metrics, and memory are created, trusted, shared, and used.05Platform StructureDefines how tools, systems, workflows, and enterprise platforms shape execution and coordination.06Governance ArchitectureDefines the standards, controls, decision rights, policies, and risk boundaries that keep the model safe.07AI AmplificationDefines how AI and agents accelerate, expose, or distort the existing operating model.
Operating Laws
The nine laws
The laws explain why the layers behave as a system. Each depends on the ones before it.
01
Structure Governs Behavior
People scale to the structure they are given. Performance problems are usually architecture problems.
Foundation law. Every other law depends on whether the operating structure is explicit.
02
Decisions Are the Atomic Unit
Organizations move through decisions. Every delay is a decision-architecture failure.
Depends on Law 1 because decision behavior follows the structure that defines it.
03
Incentives Override Intent
Declared values never beat lived incentives. Behavior follows what is rewarded.
Depends on Law 1 because incentives are part of the operating structure.
04
Information Quality Determines Intelligence
Signal clarity determines decision quality, human or AI. AI on ungoverned information amplifies disorder.
Depends on Laws 1 and 2 because information quality follows ownership and decision discipline.
05
Governance Must Reduce Entropy
Real governance creates operational clarity. Compliance theater creates the appearance of control.
Depends on Laws 1 through 4 because governance must govern real behavior, decisions, incentives, and information.
06
Architecture Precedes AI
AI amplifies the operating system beneath it. Structure before velocity.
Depends on Laws 1 through 5 because AI requires structure, decisions, incentives, information, and governance.
07
Leaders Model the System
The operating model mirrors leadership behavior. No exception.
Sustains Laws 1 through 6 because leaders make the architecture credible in practice.
08
Autonomy Is Earned
No agent or decision type advances in autonomy by default. Advancement requires clearing a measurable confidence threshold owned by the business, not engineering.
Depends on Laws 5 and 6 because governance and architecture must exist before autonomy can be safely extended.
09
Oversight Is Finite
Every governed agent consumes a named human supervisor's finite control capacity. Exceeding Supervisory Control Capacity is an operational failure, not a warning.
Depends on Laws 1 and 8 because supervision is structure and earned autonomy keeps supervisory load survivable.
Leadership
The four executive questions
The framework exists to answer these four questions. Every layer traces back to them.
2
How are decisions made?
3
How does information flow?
4
Where does AI assist versus decide?
Maturity
The five maturity phases
Sequenced, not scored in isolation. The weakest foundation caps what is safe to automate.
Phase 1
Decision Clarity
Define who owns what and how decisions are made.
Phase 2
Communication Discipline
Structure how information flows and build information governance.
Phase 3
Governance Embedding
Move controls from documents into the workflows where work happens.
Phase 4
AI Amplification
Scale AI across governed decision types.
Phase 5
Human-Machine Orchestration
The architecture self-maintains through continuous optimization.
Autonomy
The autonomy spectrum
Autonomy is earned by clearing a measurable confidence threshold set by the business owner, not engineering. Every new decision type starts at Agent Assisted.
- 1Human OnlyA human makes and executes the decision. No agent participates.
- 2Human ApprovedAn agent proposes, but a named human must approve before anything takes effect.
- 3Agent AssistedThe default entry point for every new agent or decision type. The agent drafts or recommends; a human owns the decision and the outcome.
- 4Agent SupervisedThe agent acts inside an approved corridor while a named human monitors performance, exceptions, and rollback conditions.
- 5Agent AutonomousThe agent acts and adapts inside a governed corridor with continuous monitoring and escalation. A human still owns the outcome.
Ownership
Ownership modes
The ownership mode is chosen before deployment, never discovered after.
Human in the Loop
HITLDefaultA human reviews and approves each action before it takes effect. The default for any new or high-stakes decision type.
Human on the Loop
HOTLThe agent acts, and a human monitors and can intervene or halt at any point.
Supervised Automation
The agent runs a defined process end to end while a named human owns the exceptions, the audit, and the outcome.
AI-Assisted Human Decision
A human makes the decision with the agent supplying analysis, options, and evidence.
Vocabulary
Core vocabulary
The terms LPM uses precisely. Definitions here are canonical.
- Governed agents
- Agents operated inside named human ownership, decision rights, boundaries, and audit, never as unaccountable actors.
- Knowledge Objects
- Reusable, versioned operating-model assets authored once and rendered across education, knowledge objects, assessments, and the platform.
- Supervisory Control Capacity
- The finite amount of agent oversight a single named human supervisor can carry before control degrades.
- Control Readiness Index
- A composite score of how ready the operating model is to govern and safely automate work across the seven layers.
- Lineage
- The traceable record of who decided what, on what evidence, and what changed as a result.
- Confidence gates
- Measurable thresholds a decision type must clear, set by the business owner, before it earns more autonomy.
For the full alphabetical list of defined terms, see the Glossary.
Download
Download the specification
The written specification above is the canonical source today.
Notification registration is not open. This page will expose the public download when the PDF is available.