Technology

Decision Intelligence Architecture

Autonomy is built as a Decision Intelligence Platform following Gartner's DIP framework (inaugural MQ, January 2026). The architecture implements all four DI lifecycle capabilities natively for supply chain:

• Decision Modeling — Five decision elements define every decision structure
• Decision Orchestration — Agent coordination + Authorization Protocol coordinate execution across agents
• Decision Monitoring — Drift detection, calibrated confidence, and forecast quality scoring track outcomes
• Decision Governance — Override effectiveness tracking, confidence bounds, and escalation arbiter

Three foundational decision science frameworks inform the architecture: Google's former Chief Decision Scientist provides the unifying definition — DI as the discipline of turning information into better actions. Causal Decision Diagrams (CDDs) map decision levers through intermediaries to outcomes. A sequential decision framework provides the mathematical foundation for decisions under uncertainty.

Sequential Decision Framework

A unified sequential decision framework provides the theoretical foundation for Autonomy's decision architecture. The framework defines five core elements — State, Decision, Exogenous Information, Transition Function, Objective — and four policy classes that map directly to our three-tier AI:

• Rule-based policies — Direct state-to-action mapping. The deterministic engine's base-stock rules.
• Parameterized optimization — Strategic analysis computes policy parameters for network-wide coordination.
• Learned value functions — Execution agents learn state values from outcomes via reinforcement learning.
• Simulation-based lookahead — Monte Carlo simulation for scenario evaluation and planning.

Execution Agent Architecture

Based on peer-reviewed research in compact recursive architectures, each agent uses a compact recursive architecture that achieves 90-95% of optimal policy performance while running at <10ms per decision.

The key insight: recursion multiplies compute without multiplying parameters. Rather than scaling model size (which leads to memorization), recursion forces the model to learn generalizable rules. Compact recursive models outperform models thousands of times their size on structured reasoning tasks.

Confidence Guarantees

Every AI decision carries a distribution-free confidence guarantee via a distribution-free confidence framework. Unlike heuristic confidence scores, this approach provides mathematically calibrated coverage: "P(loss > threshold) ≤ α" with guaranteed validity.

Confidence calibration runs hourly from historical decision-outcome pairs. When confidence drops below the escalation threshold, decisions automatically route to higher reasoning tiers.

Agentic Authorization Protocol

When an agent needs to take an action outside its authority boundary (cross-functional trade-offs, resource sharing, priority overrides), the Agentic Authorization Protocol (AAP) provides machine-speed negotiation. Each agent evaluates the full balanced scorecard impact of proposed actions, then requests authorization from affected agents.

AWS Supply Chain Data Model

100% compliance with 35 AWS Supply Chain entities. This ensures interoperability with enterprise ERP systems (SAP S/4HANA, APO, ECC) via RFC, OData, and CSV integration. No proprietary data formats — your data stays portable.

Escalation Arbiter

Inspired by dual-process cognitive theory, the Escalation Arbiter monitors agent decision quality over time. When persistent drift is detected (not one-off outliers, but sustained performance degradation), it routes to operational replanning or strategic replanning automatically.

Gartner SCOR Metrics Hierarchy

Autonomy implements the industry-standard SCOR (Supply Chain Operations Reference) metrics framework across three tiers that cascade from strategic assessment to operational correction:

1. Assess (Strategic) — Executive health check. Revenue growth, EBIT margin, return on capital employed. Evaluated by the strategic layer at weekly/monthly cadence.
2. Diagnose (Tactical) — Cash-flow diagnostics. Inventory turns, cash-to-cash cycle, OTIF. Evaluated by the coordination layer at daily cadence.
3. Correct (Operational) — Root cause and action. Per-agent execution metrics: decision accuracy, throughput, override rate, touchless rate. Evaluated by 11 execution agents at per-decision cadence.

Each tier maps to SCOR performance attributes — Reliability, Responsiveness, Agility, Cost, Asset Management — ensuring metrics speak the same language as existing supply chain governance frameworks.