Industrial spatial decisions are often scattered across systems that lack a common reference frame or ownership, forcing analysts to spend most of their effort reconciling data rather than reasoning about space. This paper introduces ANASSA (Agentic Network for Spatial AI Orchestration Architecture), a reference architecture that integrates structured reasoning, multi‑agent workflow orchestration, authoritative spatial grounding, and feedback‑driven execution, and demonstrates its deployment in a live supply‑chain environment.
Key components of ANASSA include:
- Typed Data Contracts at every agent boundary to explicitly define input and output schemas;
- Validation of all spatial claims against a Certified Authoritative Foundation, guaranteeing consistency across heterogeneous sources;
- Two Human‑in‑the‑Loop blocking approvals embedded within a six‑step cognitive loop, providing auditability and safety.
The deployed system handles multi‑hazard risk assessment for an ocean‑and‑inland freight lane and cold‑chain excursion detection using container telemetry. Over a three‑month proof‑of‑concept (POC) with three implementation iterations, the following outcomes were observed:
- Data quality improved by 25%;
- Spatial workflow runtime reduced by 50%;
- Decision latency decreased by 60% compared with the prior manual process.
These percentages reflect uncontrolled deployment observations rather than results from controlled experiments. To enable mechanism‑level evaluation, the paper also specifies Instrumentation metrics and Ablation Protocols for systematic analysis.
Review