TAG ARCHIVE
agentic-company
43 MARIA OS blog articles tagged agentic-company. Research on human-agent organizations, delegation boundaries, role topology, and governed autonomy. This canonical topic archive supports search engines and LLM retrieval.
Judgment OS / Decision Intelligence OS
Core MARIA OS research on turning organizational judgment into executable decision systems.
Agentic Company Architecture
Research on human-agent organizations, delegation boundaries, role topology, and governed autonomy.
Responsibility Gates and AI Governance
Safety, accountability, fail-closed gates, auditability, and human-in-the-loop control for AI agents.
Multi-Agent Mathematics
Formal models for convergence, stability, game theory, graph dynamics, and multi-agent evaluation.
Evidence, RAG, and Knowledge Governance
Evidence bundles, retrieval architecture, Graph RAG, knowledge trust, and auditable reasoning pipelines.
Agentic R&D and Judgment Science
Research operations, simulation labs, judgment science, recursive improvement, and experimental AI governance.
Turning the Founder's Mind into a Staircase Others Can See
A MARIA OS bridge theory for translating high-abstraction thinking into an intermediate language that enterprise customers, technical leads, investors, and engineering candidates can climb
Concepts like MARIA OS, Decision OS, CEO Clone, Agent Company, harness, envelope, and reflex look impressive in isolation, but depending on the listener, they easily lose their footing for understanding. This article lays out how to externalize the abstraction hierarchy inside the founder's head — not by lowering it, but as a staircase of principles, physical analogies, concrete examples, and implementation evidence. The goal is to create entry points where customers, CTOs, investors, and engineering candidates can each step in, without diluting the thinking itself.
Dynamic Harness and Phase-Space Control: From virtual-talent to MARIA OS
Reframing runtime episodes, failure taxonomies, dynamic scorecards, repair proposals, and controlled self-healing as phase control for agentic society
The central question for agentic systems is shifting from model intelligence to runtime phase control. This article defines the Dynamic Harness as a Runtime Governance Layer that observes, evaluates, and controls the phase space of an agent runtime, connecting MARIA OS research with implementation lessons from bonginkan/virtual-talent.
From AI Office to Agent HR OS: The Operating Stack for Human + AI Organizations
Why AI Office, AI Office Building, and Agent HR OS should be understood as one connected system for operating AI employees, not just using AI tools
Enterprise AI is moving from isolated assistants to managed AI labor. This article explains how AI Office provides the workplace layer, AI Office Building provides organizational topology, and Agent HR OS provides the HR and governance layer for recruiting, evaluating, promoting, and operating AI employees inside a Human + AI Organization.
How Agent Office Replaces White-Collar Execution: Workflow Transfer, Organizational Redesign, and a Staged Change Roadmap
Why the real shift is not job-title extinction but the transfer of drafting, coordination, reporting, and repeatable execution into an agent operating layer
Agent Office does not first replace white-collar employees as a category. It first replaces the hidden execution layer inside white-collar work: drafting, routing, follow-up, reconciliation, reporting, and first-pass judgment. This article uses current evidence from OpenAI, OECD, ILO, Anthropic, WEF, and NIST to model which workflows move first, how fast the shift can happen, and what a practical change-management roadmap looks like.
Command-less AI Architecture: Goal-Driven Agents That Generate Their Own Tools Without Pre-Defined Commands
Eliminating the command registry in favor of goal decomposition, plan generation, and dynamic tool synthesis
Traditional agent architectures bind agents to pre-defined command sets — fixed APIs, registered tools, and enumerated actions. This paper presents the MARIA OS command-less architecture, where agents receive goals rather than commands, decompose them into hierarchical plans, detect capability gaps, and synthesize whatever tools are needed for execution. We formalize the morphisms between Goal space G, Plan space P, and Tool space T, prove convergence of the tool space under recursive planning, and demonstrate that command-less agents achieve 3.2x higher task completion rates on novel problem classes compared to command-bound architectures.
Capability Gap Detection: The Metacognitive Layer That Enables Self-Extending Agents
How agents recognize what they cannot do and trigger autonomous self-extension through formal gap analysis
Self-extending agents require a prerequisite that most architectures ignore: the ability to know what they do not know. This paper formalizes capability gap detection as a metacognitive layer that compares required capabilities against the agent's capability model, classifies detected gaps, prioritizes them by urgency and impact, and decides whether to synthesize, request, delegate, or escalate. We introduce the capability coverage metric, gap entropy measure, and multi-agent gap negotiation protocol. Experimental results show that agents with formal gap detection achieve 4.1x fewer silent failures and 2.8x faster self-extension compared to agents relying on runtime error detection.
Self-Modifying Agent Systems: Architecture for Agents That Rewrite Their Own Tools, Commands, and Workflows
Beyond tool creation — a formal framework for bounded self-modification with stability guarantees and immutable audit trails
Agents that merely create new tools hit a ceiling. Real operational autonomy requires agents that can modify existing tools, rewrite commands, and restructure workflows based on performance feedback. We present a formal architecture for bounded self-modification with Lyapunov stability analysis, halting guarantees, and responsibility-gated audit trails.
Agent Tool Compiler: From Natural Language Intent to Executable Tool Code via Compilation Pipeline
Agents as compilers — a formal framework mapping NL intent through intermediate representation to optimized, type-safe runtime tools
Tool-generating agents are ad-hoc code producers. We reframe tool synthesis as a compilation problem: natural language intent is parsed into an Intent AST, lowered to a Tool IR (intermediate representation), optimized through security hardening and dead code elimination passes, and emitted as type-safe executable code that hot-loads into the agent runtime. This paper presents the Agent Tool Compiler architecture with formal language theory foundations.
Self-Extending Agent Architecture: Capability Gap Detection, Tool Synthesis, and Autonomous Evolution Under Governance Constraints
Agents that recognize their own limitations and autonomously build the tools they need — within the safety boundaries of an operating system
Traditional AI agents are bounded by the tools humans provide. When an agent encounters a task outside its toolset, it halts and waits. This paper introduces the Self-Extending Agent Architecture (SEAA), where agents detect their own capability gaps, synthesize new tools through code generation, validate those tools in sandboxed environments, and register them into the OS runtime — all under human-governed safety constraints. We formalize the agent state model X_t = (C, T, M, R), derive the self-extension equation X_{t+1} = E_t ∘ G_t ∘ J_t(X_t), prove Capability Monotonicity under validation gates, and demonstrate the architecture within MARIA OS's hierarchical coordinate system.
Agents That Write Their Own Tools: A 4-Phase Architecture for Tool Discovery, Synthesis, Validation, and Registration in Autonomous Systems
From static tool chains to self-extending capability — how MARIA OS agents create the tools they need at runtime
Normal agents wait for humans to build tools. MARIA OS agents create their own. This paper details the 4-phase tool lifecycle — Discovery, Synthesis, Validation, Registration — that enables agents to identify missing capabilities, generate tool implementations, verify correctness and safety in sandboxed environments, and hot-load new tools into the OS runtime. We formalize tool generation rate, quality convergence, and multi-agent tool sharing, and present a case study of an Audit agent creating an OCR extraction tool at runtime.
Agent Capability OS: Command Registry, Tool Registry, and Capability Graph as the Three Pillars of Self-Extending Agent Architecture
Why individual agents cannot manage organizational capability — and how an OS-level abstraction solves the coordination problem
As agentic organizations scale beyond dozens of agents, managing capabilities becomes a systems-level challenge that no single agent can solve. This paper introduces the Agent Capability OS — an operating system abstraction that governs how capabilities are registered, discovered, allocated, and evolved across an agent population. We formalize three core registries (Command, Tool, Capability Graph) and prove that OS-level capability management achieves O(log N) discovery latency versus O(N^2) in decentralized approaches. A case study of a 54-agent audit office demonstrates how the Capability OS manages 200+ tools across 6 organizational floors while maintaining zero capability conflicts.
Tool Genesis Under Governance: How to Safely Turn Generated Code into New Commands
A formal framework for sandbox verification, permission escalation, audit trails, and rollback mechanisms that enable self-extending agent systems without sacrificing safety
When an AI agent generates code that could become a new command in a production system, every line of that code becomes an attack surface. Without governance gates between generation and registration, a self-extending agent is indistinguishable from a self-propagating vulnerability. This paper presents the MARIA OS Tool Genesis Framework: a 7-stage pipeline that transforms generated code into governed commands through sandbox verification, formal safety proofs, permission escalation models, immutable audit trails, and automatic rollback mechanisms. We formalize tool safety as a decidable property under bounded execution, derive permission escalation bounds using lattice theory, introduce the Tool Safety Index (TSI) as a composite metric, and demonstrate that governed tool genesis achieves 99.7% safety compliance with only 12% latency overhead compared to ungoverned registration. The central thesis: self-extension is not dangerous — ungoverned self-extension is.
MARIA OS Evaluation Harness: A Standard Testing Infrastructure for Measuring Agent Quality
Formal test categories, composite scoring, and continuous evaluation pipelines that transform agent quality from subjective assessment into reproducible engineering measurement
Agent quality cannot be managed if it cannot be measured. Traditional software testing verifies deterministic input-output mappings, but AI agents operate in stochastic, multi-step decision spaces where correctness is contextual, safety is probabilistic, and governance compliance is structural. This paper introduces the MARIA OS Evaluation Harness — a standardized testing infrastructure that defines four test categories (correctness, safety, performance, governance compliance), four primary metrics (decision accuracy, gate compliance rate, evidence quality score, latency under load), and a formal composite scoring framework. We present the harness architecture comprising a test runner, scenario generator, oracle comparator, and regression detector, all scoped through MARIA coordinates for hierarchical test targeting. We prove that the composite agent score is monotonically responsive to genuine quality improvements and demonstrate that continuous evaluation pipelines catch 94.7% of quality regressions before production deployment.
Governance Load Testing: Where Does Governance Break in the 1000-Agent Era?
Stress-testing decision pipelines, approval queues, gate evaluation, and conflict detection under extreme agent concurrency to identify governance breaking points and mitigation architectures
Governance architectures designed for 10-agent teams do not survive contact with 1000 concurrent agents. Decision pipeline throughput saturates, approval queues grow unbounded, gate evaluation latency exceeds SLA windows, and conflict detection explodes as O(n^2) pairwise comparisons overwhelm detection infrastructure. This paper presents a rigorous load-testing methodology for AI governance systems, identifies precise breaking points across the MARIA OS decision pipeline, models governance bottlenecks using formal queueing theory (M/M/c and M/G/1 models), and proposes mitigation strategies including hierarchical delegation, batch approval, predictive gating, and zone-scoped conflict partitioning. We report benchmark results at 10, 100, 1000, and 10000 agent scales, demonstrating that naive governance collapses at approximately 340 concurrent agents under default configuration, while the optimized architecture sustains governance integrity up to 12000 agents with sub-second gate latency.
AI Office Operating Model: Design Principles for a Virtual Office Where 10 Teams Work as a Unified Organizational OS
Formalizing the virtual office as a graph-theoretic operating system with inter-team protocols, shared resource management, and graduated autonomy boundaries
This paper presents a comprehensive architecture for a virtual AI office where 10 specialized teams — Sales, Audit, Dev, HR, Legal, Finance, Strategy, Support, QA, and R&D — operate as a unified organizational OS. We formalize inter-team communication protocols as message-passing on a directed graph, define shared resource management through capacity allocation tensors, establish team autonomy boundaries via responsibility cones, and map the entire office to the MARIA coordinate system. The model introduces meeting scheduling agents, knowledge sharing infrastructure, team performance metrics, and conflict resolution mechanisms grounded in organizational graph theory. We prove that office-level governance and team-level autonomy can coexist under a hierarchical gate structure, achieving 89% autonomous operation while preserving 100% accountability traceability.
CEO Clone as Decision Interface: Persona Layer Design for Delegating Executive Judgment
A formal architecture for encoding executive cognition into an auditable, drift-resistant persona layer that delegates judgment while preserving principal authority
Executive judgment is the highest-leverage bottleneck in any organization. Every strategic decision that waits for the CEO creates queue delay across the entire enterprise. Yet delegation through human hierarchies introduces information loss, preference distortion, and accountability diffusion. This paper presents the CEO Clone — not a chatbot that mimics speech patterns, but a computational decision interface that encodes the CEO's values, risk tolerance, decision patterns, and communication style into a formally verifiable persona layer. We model judgment delegation as a principal-agent problem with information asymmetry, introduce decision fidelity metrics with drift detection, and design calibration loops that maintain clone-principal alignment over time. The architecture operates within MARIA OS governance infrastructure, ensuring every delegated decision produces an immutable audit trail with full traceability to the encoded persona parameters that produced it.
Audit Universe Runtime: Agent Design for Executing Audit Procedures as Runtime Operations
Transforming ISA/JICPA standards into executable agent specifications — from sampling strategies to substantive testing, within a MARIA OS governance architecture
Traditional audit procedures are encoded in prose-based standards that resist automation. This paper presents the Audit Universe Runtime — a multi-agent execution environment within MARIA OS that compiles audit standards (ISA, JICPA) into executable agent task specifications. We formalize audit procedures as state machines, design sampling strategy agents with statistical rigor, implement real-time anomaly detection during substantive testing, and prove audit completeness through a formal coverage model. The architecture maps MARIA coordinates to engagement structures, enabling continuous auditing with immutable audit trails and human-agent collaboration gates at every materiality threshold.
MARIA OS Appliance Reference Architecture: Standard Configuration for On-Premise AI Governance Infrastructure
A complete hardware and software blueprint for deploying MARIA OS as a self-contained appliance — covering GPU/CPU sizing, network topology, security hardening, HA clustering, disaster recovery, and TCO analysis for regulated enterprises
Cloud-native AI platforms dominate the conversation, but regulated industries — finance, healthcare, defense, critical infrastructure — face a hard constraint: sensitive decision data cannot leave the building. This reference architecture defines the MARIA OS Appliance: a rack-mountable, air-gap-capable governance platform that runs the full multi-agent decision pipeline on-premise. We specify hardware tiers from single-node evaluation units to multi-site federated clusters, detail the software stack from OS kernel to agent runtime, prove that governance guarantees hold under network partition, and provide a TCO framework that quantifies the break-even point against cloud deployment. The result is a turnkey AI governance infrastructure that preserves data sovereignty without sacrificing capability.
Agentic Ethics Lab: Designing a Corporate Research Institute for Structural Ethics in AI Governance
A four-division, gate-governed research architecture that transforms ethics from philosophical declaration into executable, auditable, and evolvable system infrastructure
Ethics declarations without structural enforcement are organizational theater. This paper presents the Agentic Ethics Lab — a corporate research institute embedded within the MARIA OS governance architecture, operating as a first-class Universe with four specialized divisions: Ethics Formalization, Ethical Learning, Agentic Company Design, and Governance & Adoption. Each division runs agent-human hybrid teams under fail-closed research gates. We formalize the lab's architecture using decision graph theory, prove that self-referential governance research preserves safety invariants, and demonstrate that a corporate research institute with no revenue targets but strategic alignment outperforms both pure academic and pure product research in responsible AI advancement.
Cross-Domain Research Governance: A 12-Month Integrated Research Plan for Capital, Operational, and Physical AI Systems
Orchestrating four parallel research streams across capital decision engines, operational agentic companies, robot judgment systems, and holding integration under unified gate governance
Research programs that operate in isolation produce findings that cannot be integrated. Capital decision engines optimized without operational context misallocate resources. Operational agentic companies designed without capital awareness cannot sustain themselves. Robot judgment systems built without holding-level governance create liability gaps. This paper presents a 12-month cross-domain research plan for an Autonomous Industrial Holding that integrates four parallel streams — Capital Decision Engine (Stream A), Operational Agentic Company (Stream B), Robot Judgment OS (Stream C), and Holding Integration (Stream D) — under unified research gate governance. We formalize stream dependency graphs, derive milestone probability models using PERT/CPM analysis, introduce cross-stream conflict detection metrics, model research velocity and throughput, express gate passage probability as a function of research maturity, and quantify integration risk propagation across streams. The plan covers 20 research themes (4 streams x 5 themes each) with detailed experiment designs, statistical methodology, and KPI specifications. Research gates RG0-RG3 govern all outputs with fail-closed semantics. The central thesis: cross-domain research governance is not project management — it is a decision architecture problem that requires the same structural rigor as the systems it studies.
Decision Civilization Infrastructure: From Ethics-as-Architecture to the Universal Responsibility Operating System
The capstone synthesis — why the AGI era demands not smarter AI but better responsibility structures, and how MARIA OS unifies capital, physical, ethical, and organizational decisions under a single governance topology
Every decision an organization makes — from board strategy to robot arm trajectory, from capital allocation to ethical constraint evaluation — flows through an implicit responsibility structure. In most organizations, that structure is invisible, informal, and fragile. This paper presents the Decision Civilization Infrastructure: a unified mathematical framework that formalizes the entire decision space as a product manifold D = D_capital x D_physical x D_ethical x D_organizational, proves that responsibility is a conserved quantity under decision composition, derives scaling theorems for governance preservation as systems grow, and demonstrates that all prior MARIA OS research programs — ethics formalization, ethical learning, agentic company design, investment engines, robot judgment, responsibility decomposition, gate control theory, and quality convergence — are projections of a single underlying architecture. We introduce a category-theoretic view of decision composition across domains, establish information-theoretic bounds on decision quality, and prove convergence of all subsystems toward a stable governance attractor. The competitive moat is not AI capability but structural responsibility: mathematics, reproducibility, and fail-closed architecture that compounds over time.
Mission-Constrained Optimization in Agentic Companies
A Mathematical Framework for Value-Preserving Goal Execution
Local goal optimization often conflicts with organizational Mission. We formalize this conflict as a constrained optimization problem over a 7-dimensional Mission Value Vector, derive the alignment score and penalty-based objective, and present a three-stage decision gate architecture that prevents value erosion while preserving goal-seeking performance.
Survival Optimization and Mission Constraint Theory
Does Evolutionary Pressure Reduce Organizations to Pure Survival Machines? A Mathematical Analysis of Directed vs. Undirected Evolution
When organizations are modeled as evolutionary subjects, does the theoretical limit reduce to survival-probability maximization? This paper examines two regimes — unconstrained local optimization (λ→0) where ethics and culture are mere byproducts, and Mission-constrained optimization where evolution gains direction. We derive the survival-alignment tradeoff curve S = S₀·exp(−αD), prove Lyapunov stability of Mission erosion dynamics under dual-variable feedback control, present 7-dimensional phase diagrams for operational monitoring, and demonstrate a civilization-type phase transition where accumulated institutional improvements qualitatively change the system's risk profile.
Metacognition in Agentic Companies: Why AI Systems Must Know What They Don't Know
Latent governance density, observable metacognitive coverage, and the stability bounds of self-governing enterprises
We formalize an agentic company as a graph-augmented constrained Markov decision process G_t = (A_t, E_t, S_t, Pi_t, R_t, D_t), distinguish latent governance density D_t from observable constrained-candidate coverage D_hat_t on router-generated Top-K actions, and define damping via kappa_t = kappa(D_hat_t). The exact local contraction condition is (1 - kappa_t) lambda_max(W_t) < 1, while the buffered operating envelope lambda_max(W_t) < 1 - kappa_t preserves adaptation headroom. Governance constraints thereby function as organizational metacognition: each constraint is a point where the system observes its own behavior. Planet-100 simulations validate that buffered role specialization emerges in the intermediate governance regime.
Institutional Design for Agentic Societies: Meta-Governance Theory and AI Constitutional Frameworks
From Enterprise Governance to AI Constitutions: How Institutional Economics and Meta-Governance Theory Stabilize Multi-Agent Societies
Multi-agent AI societies require more than individual metacognition: they also require institutional design. This article formalizes agentic-company governance, derives social objective functions for AI-human ecosystems, establishes the Speed Alignment Principle as a stability condition, and presents an AI-constitution model with revision rules. In simulations across 600 runs, adaptive institutional frameworks reduced spectral radius from 1.14 to 0.82 while maintaining audit scores above 0.85.
Meta-Insight Under Distribution Shift: Change-Point Governance Loops for Enterprise Agentic Systems
An operational architecture for detecting non-stationarity, throttling unsafe adaptation, and restoring decision quality under drift
This article outlines change-point detection, bounded policy updates, and fail-closed escalation for distribution-shift governance.
Counterfactual Escalation Policy: Meta-Insight Routing for High-Impact Human Review
Estimate intervention value before handoff to reduce unsafe approvals and unnecessary escalations
Escalation is triggered when estimated causal benefit exceeds review cost, not by confidence alone.
Memory Stratification for AI Governance: A Rate-Distortion Framework for Retention Decisions
Use information theory to decide what enterprise AI systems should remember, summarize, or discard
Rate-distortion memory policy retains high-utility context while limiting latency, privacy risk, and contradiction noise.
Securing Recursive AI Feedback Loops: Adversarial Reflexivity Hardening for Meta-Insight Systems
Defense framework for prompt injection, feedback poisoning, and policy-hijack attacks in self-improving loops
Layered provenance checks, anomaly scoring, and quarantine rules harden adaptive loops while preserving auditability.
Causal Analysis of Organizational Learning Rate: OLR Decomposition for Intervention Attribution
From correlation-heavy dashboards to intervention-level attribution in meta-insight governance systems
Causal OLR decomposition attributes observed learning-rate gains to specific interventions, improving budget and policy allocation decisions.
Causal-Temporal Knowledge Graph for AI Governance: Path-Specific Responsibility Attribution
A deep research framework for path-specific accountability, time-aware causality, and audit-grade explanation in enterprise AI
A temporal responsibility graph enables path-level causal attribution and faster, more reproducible root-cause analysis.
Governing Emergent Role Specialization: Stability Laws for Agentic Companies Under Constraint Density
A mathematical framework for calibrating governance in self-organizing enterprises
We distinguish the exact contraction condition `(1 - D) · λ_max(A) < 1` from the conservative operating envelope `λ_max(A) < 1 - D`, giving enterprise architects a rigorous way to tune governance density in agentic organizations.
The Algorithm Stack for Agentic Organizations: 10 Essential Algorithms Mapped to a 7-Layer Architecture
Beyond generative AI: a practical computational substrate for self-governing enterprises
An agentic company is not built on generative AI alone. We present 10 core algorithms across language, tabular prediction, state-transition control, graph structure, and anomaly detection, organized into a 7-layer architecture for enterprise governance workloads.
Transformer Architecture for Agentic Language Intelligence: Self-Attention as the Cognitive Layer of Enterprise Decision Systems
How self-attention enables multi-agent context fusion, decision-log comprehension, and hierarchical organizational reasoning
Transformer architectures are central to enterprise language understanding, but production decision systems require additional design constraints. This paper formalizes transformers as the Cognition Layer (Layer 1) of the agentic company stack, introduces cross-agent attention for organizational context fusion, adapts positional encoding to hierarchical coordinates, and outlines training objectives for decision logs, contracts, meeting notes, and specification documents. In evaluated MARIA OS workloads, coordinate-aware attention reduced cross-agent context fusion error by 34% versus standard multi-head attention, and hierarchical positional encoding improved organizational structure extraction F1 by 28%.
Gradient Boosting for Enterprise Decision Prediction: XGBoost and LightGBM as the Decision Layer of Agentic Companies
Why enterprise data is often tabular and how gradient boosting ensembles support approval prediction, risk scoring, and outcome estimation
While deep learning dominates many unstructured tasks, enterprise decision data is frequently tabular: structured features describing decisions, agents, contexts, and outcomes. This paper formalizes gradient boosting (XGBoost/LightGBM) as the Decision Layer (Layer 2) of the agentic company stack, details feature-engineering patterns for enterprise decision tables, and introduces SHAP-based explainability workflows for governance audits. Across evaluated datasets, the approach achieved 91.3% approval-prediction accuracy, 0.94 AUC on risk scoring, and full SHAP traceability integrated with MARIA OS responsibility gates.
Random Forest for Interpretable Organizational Decision Trees: Extracting Governance Logic from Ensemble Structure
How bagging-based tree ensembles reveal decision-branch structure, critical governance variables, and auditable policy trees
While gradient boosting often targets predictive accuracy, random forests provide a complementary strength: structural interpretability. This paper positions random forests as an interpretability engine within the Decision Layer (Layer 2), showing how ensemble structure surfaces governance logic, highlights key variables through permutation/impurity importance, and yields auditable policy trees. In evaluated workloads, random-forest feature importance reached 0.93 rank correlation with domain-expert rankings, extracted trees matched 89% of documented governance policies, and out-of-bag error supported validation in data-constrained settings.
Markov Decision Processes for Business Workflow State Control: Formalizing the Agentic Company as a State Transition System
How MDPs, Bellman equations, and policy optimization support workflow control, responsibility decomposition, and gate-constrained automation
The agentic company can be modeled as a state-transition system. Business workflows move through discrete states — proposed, validated, approved, executed, completed — with transitions governed by policies balancing efficiency, risk, and human authority. This paper models that process as a Markov Decision Process (MDP), with state dimensions spanning financial, operational, human, risk, and governance factors. We derive Bellman equations for policy optimization, analyze gate-constrained MDP behavior when specific transitions require human approval, and map the MARIA OS decision pipeline to a finite-horizon MDP with responsibility constraints. In tested workflow graphs, policy iteration converged within 12 iterations and yielded 23% throughput improvement over heuristic routing while keeping governance compliance at 100%.
Actor-Critic Reinforcement Learning for Gated Autonomy: PPO-Based Policy Optimization Under Responsibility Constraints
How Proximal Policy Optimization enables medium-risk task automation while respecting human approval gates
Gated autonomy requires reinforcement learning that respects responsibility boundaries. This paper positions actor-critic methods — specifically PPO — as a core algorithm in the Control Layer, showing how the actor learns policies, the critic estimates state value, and responsibility gates constrain the action space dynamically. We derive a gate-constrained policy-gradient formulation, analyze PPO clipping behavior under trust-region constraints, and model human-in-the-loop approval as part of environment dynamics.
Multi-Armed Bandits for Enterprise Strategy Optimization: Thompson Sampling, UCB, and Contextual Bandits in Agentic Organizations
How exploration-exploitation algorithms form the fifth layer of the agentic company architecture
Enterprises continually face the exploration-exploitation dilemma: whether to exploit known strategies or test potentially better alternatives. This paper formalizes multi-armed bandits as the Exploration Layer (Layer 5), covering Thompson sampling with Beta priors, UCB confidence bounds, contextual bandits for personalized decisions, and Bayesian optimization for business hyperparameter tuning. We provide enterprise-oriented regret analysis and describe integration with the MARIA OS strategy engine.
Graph Neural Networks for Organizational Network Dynamics: Message-Passing, Spectral Convolutions, and Influence Propagation in Agentic Hierarchies
How GNNs form the Structure Layer that models agent dependencies, information flow, and hierarchical topology in self-governing enterprises
Agentic companies can be modeled as graph structures, where agents connect through dependencies, information channels, and approval chains. This paper formalizes Graph Neural Networks as the Structure Layer (Layer 3), covering message-passing networks for organizational flow, spectral convolutions for hierarchy discovery, graph attention for dynamic topology, and link prediction for emerging dependencies. We also analyze influence-propagation matrices and spectral-radius indicators for governance stability, and describe integration with the MARIA OS Universe visualization.
Clustering Algorithms for Emergent Agent Role Specialization
How k-means, DBSCAN, and hierarchical clustering form the computational mechanism of organizational role formation
Role specialization in agentic companies can be analyzed as a clustering phenomenon. We show how k-means supports initial role assignment, DBSCAN discovers natural clusters without fixed role counts, and hierarchical clustering models nested organizational structure. We derive a role-specialization equation and describe how MARIA OS applies dynamic re-clustering for organizational adaptation.
Anomaly Detection for Agentic System Safety and Deviation Control
Isolation Forest and Autoencoder reconstruction error as the computational safety layer for self-governing enterprises
Agentic systems can produce operational deviations that require early detection and controlled response. This paper combines Isolation Forest anomaly scoring with Autoencoder reconstruction error to build a layered safety monitor. We define an anomaly-throttle-freeze response cascade and show how the MARIA OS stability guard applies the spectral-radius condition `spectral_radius < 1 - governance_density` in runtime governance.
Agentic Company Structural Design: Responsibility Topology, Conflict-Driven Learning, and Self-Evolving Governance for Human-Agent Organizations
Modeling the enterprise as a responsibility topology across human-agent decision nodes
This paper explores corporate design where the primary unit is the decision node and its responsibility allocation, not only role or department labels. It introduces five linked research programs that model the enterprise as a weighted directed responsibility graph whose topology evolves through conflict-driven learning. We formalize human-agent responsibility matrices, derive scalable topology conditions, define health metrics for hybrid organizations, and model governance as a self-evolving decision graph with gate-managed policy transitions.