Glossary

Agent Fleet Engineering is the discipline of designing, deploying, and operating coordinated multi-agent systems for enterprise workflows. Where Robotic Process Automation (RPA) handles a single rule-based task and generic AI agents handle isolated queries, Agent Fleet Engineering deploys structured groups of specialized agents that work together under orchestration, with explicit human checkpoints on decisions that carry regulatory, safety, financial, or reputational consequence.

NSigma's practice operates through the ARMOR methodology and draws on a library of 43 pre-built agents for asset-heavy operational workflows — predictive maintenance, route optimization, inspection automation, compliance documentation, anomaly triage, and more.

See also: ARMOR · MOSAIC · Agent Fleet Engineering overview

ARMOR is NSigma's five-phase deployment methodology for taking an asset-heavy operation from manual processes to live agent fleets in 90 days.

ARMOR replaces the traditional "proof-of-concept then pilot then deploy" cycle with production deployment from week one, on a defined slice of the operation.

MOSAIC stands for Managed Orchestration for Structured Analytics Intelligence Capability. It is NSigma's end-to-end analytics platform, integrating four disciplines that most organizations buy and run separately:

• dataACQUIRE — Multi-source ingestion, normalization, and storage across structured, semi-structured, and unstructured data. Handles IoT telemetry, SCADA feeds, BMS data, CRM/ERP records, market data, and unstructured documents.
• dataPREDICT — Forecasting, anomaly detection, classification, and feature engineering. Models are designed for non-linear patterns, concept drift management, and domain-specific ontologies.
• dataVISION — Computer vision on edge or cloud, powering the ViDA product and any custom vision workflow.
• dataADVANCE — Advanced analytics, optimization, simulation, causal inference, and bespoke modeling for domain-specific questions.

MOSAIC is delivered as a managed service: NSigma operates the platform; clients consume outputs (dashboards, predictions, alerts, optimized routes). This avoids the staffing, retention, and tool-sprawl cost of building the same capability in-house.

ReMI is NSigma's operating intelligence platform built on the dataACQUIRE and dataPREDICT disciplines of MOSAIC. It ingests IoT telemetry from equipment, sensors, and building systems, and turns it into:

• Predictive maintenance — Component failure prediction with 7–30 day lead time
• Anomaly detection — Real-time deviation alerts before failure
• Energy optimization — Continuous tuning of HVAC, lighting, and process loads

ReMI is protocol-agnostic (BACnet, Modbus, MQTT, OPC UA, LoRaWAN, proprietary BMS APIs) and does not replace existing BMS, SCADA, or CMMS systems — it layers on top and writes signals back.

Typical outcomes: 20–35% OpEx savings on monitored assets and 60–75% reduction in emergency reactive calls.

FRIT stands for Field Operations Intelligence and Mobile Workforce Automation. (The acronym is read "frit," rhyming with "fit.") FRIT digitizes the work that happens away from a desk: inspections, work orders, dispatch, routing, vendor coordination, and compliance documentation.

Core capabilities:

• Digitized inspections — Mobile forms with offline capture, geo-tagging, photo evidence, signature workflow, and full audit trails
• Dispatch and route optimization — Continuous job sequencing across the workforce based on skill match, SLA windows, traffic, and equipment availability
• Vendor accountability — SLA tracking, completion verification, and exception reporting for third-party crews
• Integration — Bidirectional sync with CMMS, ERP, and asset-management platforms

FRIT replaces paper inspection forms, manual dispatch boards, and fragmented work-order trails.

ViDA stands for Vision Data Analytics. It is NSigma's computer vision intelligence platform — the production form of the dataVISION discipline.

ViDA processes existing camera feeds at the edge, using models such as YOLOv8 deployed via TensorRT, to detect operational events without storing or transmitting raw video. Event classes include:

• Spills, obstructions, and PPE violations
• Crowd density and queue length
• Door events, tailgating, unauthorized access
• Parking occupancy and people counting

Because processing happens at the edge and only event metadata is retained, ViDA deployments are privacy-by-design and GDPR-compliant.

Typical capability: above 95% precision on targeted event classes after site-specific tuning.

The four integrated disciplines that make up MOSAIC.

• dataACQUIRE is the ingestion and normalization layer. ETL/ELT pipelines, real-time streaming, data lake/warehouse schemas, infrastructure-as-code deployments, and quality control across heterogeneous sources.
• dataPREDICT is the modeling layer. Forecasting with non-linear patterns, NLP with domain ontologies, feature engineering and tuning, and concept-drift management.
• dataVISION is the computer vision layer. Edge and cloud inference, model tuning per site or asset class, and event extraction without raw video retention.
• dataADVANCE is the analytics and optimization layer. Interactive dashboards, query optimization, information architecture, advanced analytics (causal inference, simulation, optimization), and WCAG 2.1 AA-compliant interfaces.

These four disciplines are intentionally non-linear and iterative — outputs from one feed back into the others over the life of a deployment.

NSigma's standard engagement model. A 90-day timeline from kickoff to live operations, structured around the ARMOR methodology. Unlike traditional proof-of-concept engagements that produce slideware or one-off models, a 90-day pilot deploys a working slice of the operation into production — with measurable outcomes defined upfront.

NSigma's catalog of 43 pre-built agents for asset-heavy operational workflows. Each agent is a composable unit handling a specific function (e.g., anomaly triage, dispatch optimization, contamination classification, compliance documentation). Agent fleets are assembled by selecting and orchestrating relevant agents during the Refine phase of ARMOR.

NSigma operates under a published AI governance framework covering model lifecycle controls, decision logging, escalation paths, human oversight requirements, bias monitoring, and audit trails. The full framework is published at /ai-governance.

Operational baseline:

• ISO 27001 certified
• GDPR-compliant deployments where applicable
• SOC 2 alignment available on request
• All material decisions traceable to a logged inference + human approval (where applicable)

NSigma's term for industries where the operating model is dominated by physical assets (buildings, fleets, grids, equipment) rather than purely digital products. The four core verticals served are:

• Commercial Real Estate — owners, operators, REITs
• Utilities — electric, water, gas
• Waste Management — haulers, transfer stations, MRFs
• Investment Management — asset managers, hedge funds (assets here refers to AUM and data assets rather than physical assets, but the operating model — large, heterogeneous data systems and high cost of fragmentation — shares the same structural challenges)