ARKA AI Glossary

Indefensible Evidence 🔗

The flagship outcome of the ARKA Sovereign Platform. Notarized, tamper-evident proof that a decision was made according to policy and authorized by the correct authority. Designed to be incontrovertible in audits, legal reviews, and commercial reconciliations.

Outcome Verification (Sovereign Ledger) 🔗

The runtime process of generating **Evidence Bundles**. Every decision cycle is sealed (hash + signature) and recorded in the Sovereign Evidence Ledger. Verification ensures that all automated actions remain attributable and defensible.

Policy Gate 🔗

Runtime policy enforcement layer that validates all decisions against customer-defined rules before execution. Policy packs contain customer-defined governance rules. Policy bindings connect packs to missions. Policies are evaluated at runtime, not post-hoc.

Avoid calling it: "rule engine" (too generic), "validator" (incomplete)

Authority Router 🔗

Control plane component that routes decisions to appropriate approval authorities based on impact thresholds, risk levels, and organizational hierarchy. Uses the Authority Matrix to determine routing rules.

Decision Fabric 🔗

ARKA AI's deterministic execution spine (formerly Decision Fabric) coordinating policy, authority, and evidence. The Kernel ensures that no action occurs without a binding policy check and notarized verification. It is the technological enforcement of organizational sovereignty.

Mission Control 🔗

Governed execution UI for mission setup, monitoring, approval workflows, and evidence review. The primary interface for operators and executives. Includes Studio (build interface), monitoring dashboards, and admin panels.

Outcome Navigator (ON) 🔗

Domain-specific product modules that actively steer measurable outcomes (Revenue, Compliance, Legal, Logistics) toward contracted KPI targets. Each navigator is governed by the Commercial Spine, steered by the Outcome Navigation Engine, and produces evidence-backed proof of convergence. See full ON definition.

Formerly: Outcome Engine. Modules: ON: Revenue, ON: Compliance, ON: Legal, ON: Logistics.

Sovereign Workforce (The Golden Fleet) 🔗

A registry of **Industrial Primitives** (Sovereign Workers) with deterministic responsibilities. Unlike generic agents, Sovereign Workers operate under the strict authority of the Decision Fabric and produce notarized proof of every action.

Evidence Bundle 🔗

A cryptographically sealed evidence bundle (formerly Evidence Bundle) representing a single decision cycle. Contains objective proof of Signal, Policy Validation, Authority Authorization, and Evidence. Evidence Bundles are the atomic unit of the Sovereign Evidence Ledger.

Note: Evidence produced by ARKA AI is designed to support audits and investigations; it does not replace formal certifications or regulatory determinations.

Governance Seal 🔗

Cryptographic hash computed over canonical serialized structure. Used to detect tampering. Sealing process includes: canonical serialization (deterministic JSON), hash computation, digital signature generation, and chain linking (previous hash reference).

Seal Verification 🔗

Cryptographic verification process: (1) recompute hash from canonical serialization, (2) compare with stored hash, (3) verify signature with public key, (4) check chain link to previous Evidence Bundle. Result: VALID (untampered) or INVALID (tampered/corrupted).

Revalidation 🔗

Later integrity check performed after initial Evidence Bundle creation. Designed to detect tampering and verify long-term integrity. Revalidation does not modify data—it only verifies. Frequency: on-demand or scheduled.

Execution Boundary 🔗

Runtime enforcement layer unifying Policy Gate (runtime validation), Variance Bands (acceptable operating bounds), and Human Gate (manual escalation triggers). Acts as "runtime law" designed to constrain autonomous behavior within safe operating bounds.

Trust Anchor 🔗

Durable tamper-evident ledger store used to verify evidence. Designed to support non-repudiation and long-term audit requirements. Not a —uses cryptographic sealing and chain linking for tamper detection.

Canonical Serialization 🔗

Deterministic JSON rules for hashing. Rules: sort keys alphabetically, no whitespace, UTF-8 encoding, recursive application to nested objects. Purpose: enables cryptographic verification without hash mismatches due to key ordering.

Outcome Contract 🔗

Formal specification of expected outcomes, success criteria, variance bands, and governance constraints for a mission. Defines what success looks like and acceptable operating bounds. Includes baseline declaration and commercial terms.

Variance Bands 🔗

Acceptable range for automated execution defined by policies. When decisions fall outside these bands, they trigger human approval gates. Designed to constrain autonomous behavior within safe operating bounds.

Human Gate 🔗

Approval checkpoint triggered when decisions exceed authority thresholds or fall outside variance bands. Routes to appropriate human decision-maker based on impact and risk. Execution pauses until approved or rejected.

Sovereign Sandbox 🔗

The non-production validation environment (formerly Sovereign Sandbox) where missions execute in parallel with reality. Used to validate feasibility, test policy impact, and generate "Shadow" Evidence Bundles before production promotion.

Commercial Spine 🔗

The binding architectural layer that connects technical execution (Decision Fabric) to financial outcomes (Outcome Contracts). The Spine ensures that every dollar of value delivered is backed by notarized evidence.

Promotion Gate 🔗

Governance checkpoint for moving missions from Shadow to Active execution. Requires explicit approval and validation of feasibility, policies, and variance bands. Designed to prevent premature production deployment.

Baseline Declaration 🔗

Cryptographically hashed definition of the starting point (baseline) for measuring outcome value. Once approved, it is versioned and cannot be changed without creating a new contract version. Used to calculate delivered value above baseline.

Commercial Terms 🔗

Pricing model, caps, and measurement methodology for an outcome contract. Versioned and hashed for audit integrity. Pricing is aligned to measured, evidence-backed outcomes — not seats, tokens, or usage.

Confidence Bands 🔗

Confidence tiers mapped to billing eligibility and fee scaling. Higher confidence = higher fee multiplier. The grading system ensures that ARKA AI only charges full fees when outcomes meet or exceed agreed confidence thresholds.

Billing Eligibility 🔗

Determines whether an Evidence Bundle is eligible for billing based on decision confidence scoring and commercial terms. Higher-confidence outcomes qualify for full billing; lower-confidence or exploratory outcomes have reduced or no billing eligibility.

Financial Outcome 🔗

Ledger entry linked to Evidence Bundle + contract via audit chain. Includes idempotency key (prevents duplicates), outcome value, fee value, and billing status (PENDING, APPROVED, PAID). Designed to support outcome-based billing verification.

Kill-Switch 🔗

Emergency block designed to stop execution rapidly at mission, blueprint, tenant, or global scope. Activation triggers: security incident, compliance violation, customer request, platform emergency. Effects: immediate halt, pending missions blocked, evidence sealed at interruption point.

Incident 🔗

Operational issue tracked against mission/blueprint. Types: PLATFORM_ISSUE, LOGIC_DRIFT, DATA_FAILURE, INTEGRATION_ERROR. Severity: INFO, WARNING, CRITICAL, EMERGENCY. Includes responsibility owner assignment.

Escalation 🔗

Formal routing to appropriate owner. Actions: RETRY (automatic retry with backoff), HUMAN_APPROVAL (manual approval required), KILL_SWITCH (emergency block), CUSTOMER_ESCALATION (route to customer support).

Responsibility Owner 🔗

Party responsible for incident resolution: ARKAI (platform infrastructure), CUSTOMER (data quality, business logic), SI (integration and customization), SHARED (requires collaboration).

Feasibility Analysis 🔗

Pre-execution validation that checks: data availability, system connectivity, policy completeness, and authority configuration. Designed to prevent runtime failures. Runs before Sovereign Sandbox activation.

Signal 🔗

Input data or event that triggers a decision workflow. Can be scheduled, event-driven, or manual. All signals are logged in the Evidence Ledger with timestamp and source context.

Notarization 🔗

Process of sealing Evidence Bundles: (1) cryptographic hash computation, (2) digital signature generation, (3) chain linking (previous hash reference), (4) timestamping. Creates tamper-resistant proof of execution. Links to Evidence Bundle and Governance Seal.

Secret Registry 🔗

Tenant-scoped encrypted credential store. Designed to support tenant isolation controls—credentials are strictly scoped to prevent cross-tenant access.

Identity Mapping 🔗

Connects IdP groups from your identity provider to ARKA AI authority roles. Designed to support deterministic mapping of external corporate identities to platform permissions.

Authority Matrix 🔗

Defined governance rules specifying approval limits, SLAs, and escalation behaviors per role. Used by Authority Router to determine routing decisions.

ArkLink (Integration Mesh) 🔗

Standard for tool invocation and connector safety. Standardizes how workers discover and invoke tools. Designed to support safe integration with enterprise systems (Salesforce, SAP, databases).

Native Skill 🔗

In-process capability (e.g., PDF OCR, data transformation) for performance-critical operations. Bypasses network overhead of remote tool calls for tight loops and local operations.

Sovereign Deployment 🔗

Customer-controlled infrastructure deployment (VPC or air-gapped) where all data, models, and execution remain within customer boundaries. No external dependencies for runtime operation. Designed to support data residency and compliance requirements.

Account Isolation 🔗

Architectural pattern using account_id-scoped identity, keys, and query filters designed to prevent cross-account data access. Each customer's data, policies, and execution context are strictly separated at every layer of the platform.

Connector 🔗

Integration adapter for enterprise systems (CRM, ERP, databases). Each connector includes data perimeter controls and tenant-scoped credential management designed to support safe system integration.

Outcome Navigator (ON) 🔗

ARKA AI's product module brand for domain-specific outcome steering. Each ON module actively steers business KPIs toward contracted targets through governed execution, drift detection, and convergence proof. Modules: ON: Revenue, ON: Compliance, ON: Legal, ON: Logistics.

Formerly: Outcome Engine

Deterministic KPI Steering 🔗

The steering methodology used by the NavigatorKernel to drive KPIs toward contractual targets through governed, deterministic parameter adjustments — not heuristic tuning. Uses a mathematical model of the system to predict future behavior and compute optimal control actions over a finite horizon.

Technical name: Model Predictive Control (MPC)

Outcome Navigation Engine 🔗

The core platform capability that powers all ON modules. Actively steers mission KPIs toward contractual targets using deterministic KPI steering. Encompasses the NavigatorKernel, KPI checkpoints, drift detection, and governed re-baseline proposals. Converts passive evaluation ("did we hit the target?") into active steering ("how do we converge on it?").

NavigatorKernel 🔗

The 7-engine deterministic controller at the core of the Outcome Navigation Engine. Maintains a transition model where KPI(t+1) = KPI(t) × (1+δ), with δ bounded by policy-defined limits. Supports drift detection, governed parameter adjustment, and monotonic convergence proof. All parameter changes require governance approval.

Decision Lineage 🔗

A first-class, write-once forensic record binding the complete decision path from mission inception to financial outcome. Contains six bindings: Mission + Blueprint, Worker Decisions, Governance Evaluation, Evidence Chain State, Baseline KPIs, and Financial Outcome. Answers: "Show me the full decision path that produced this outcome."

Digital Twin 🔗

A virtual model of a customer's business process or asset — enriched asynchronously with data from integrations. Used for concordance scoring (comparing AI predictions against reality), simulation-based feasibility analysis, and KPI projection. Part of the Intelligence Layer.

Decision Mining 🔗

The process of analyzing historical decision records to extract patterns, identify decision-quality trends, and surface optimization opportunities. Works against the immutable Evidence Ledger — no new data is created, only analytical queries against existing governed records.

Certified Blueprint 🔗

A pre-defined, hash-verified mission template that specifies which Sovereign Workers execute, which policies bind, and which tools are available. Blueprints follow a strict lifecycle: DRAFT → CERTIFIED → REVOKED. SHA-256 hash verification prevents unauthorized modifications (anti-fork). Only certified blueprints can execute in governed profiles.