Q: What does Cognizant mean by a “central nervous system” for factories? A: It’s a Physical AI Platform-as-a-Service that links sensors, cameras, robots and digital twins to an agentic AI layer that reasons and acts. The platform includes governance features such as audit trails and a decision ledger and is intended to work across multiple sectors including manufacturing, utilities, logistics, healthcare, and retail. Q: How should manufacturers begin when learning how to integrate industrial AI platforms? A: When learning how to integrate industrial AI platforms, start by choosing one clear business outcome and 3–5 KPIs tied to a specific line, cell, or asset with visible pain such as downtime or scrap. Lock the KPIs before selecting tools and set a 90-day target (for example, 5% more throughput or 10% fewer stops) to measure early value. Q: Which data sources and protocols should be mapped for an initial use case? A: Inventory PLCs, SCADA, MES, QMS, CMMS, historians, vision systems, robots, and ERP as primary sources and plan connectors using OPC UA, MQTT, REST APIs, and edge adapters. Create a simple data model such as an ISA‑95 asset hierarchy with ISA‑88 recipes and time-series plus events to align data to the chosen KPIs. Q: How do you decide what functions run at the edge versus in the cloud? A: Use edge computing for low‑latency control and vision tasks and use the cloud for training, benchmarking, and fleet-level insights. Plan data flow so raw signals remain at the edge, features feed the platform, and actions return to MES/CMMS/robots, while enforcing network rules like OT VLANs, telemetry brokers, and an API gateway. Q: What governance and safety controls should be in place before automating decisions? A: Build governance before automation by tracking every AI decision with audit trails and a decision ledger and by setting guardrails such as human‑in‑the‑loop gates, stop rules, and rollback steps. Write clear ethics and safety policies that specify allowed actions, prohibited behaviors, and who reviews exceptions. Q: What role do system integrators play in delivering value from industrial AI platforms? A: System integrators or skilled internal teams handle data mapping, connectors, workflow design, and change management that tie platforms to real operations. Select integrators with vendor‑agnostic skills, OT security experience, MES/CMMS knowledge, and proof of scaled go‑lives, and agree up front on ownership of PLC tags, data models, and agent sign‑offs. Q: What is the recommended pilot approach and timeline for scaling site by site? A: Pilot one use case on a single line for 8–12 weeks, freeze the scope, measure weekly against the KPIs, and decide to kill or scale based on the KPI trend. When successful, document a playbook with connectors, data models, SOPs, training, and rollback plans to replicate at the next site. Q: What common pitfalls should teams avoid and how can they measure ROI quickly? A: Avoid common pitfalls like trying to boil the ocean, weak data contracts, shadow point‑to‑point integrations, unclear authority for overrides, and metrics drift by locking baselines and using control charts. Measure ROI by tying KPIs to money—throughput, scrap, labor, energy, and spare parts—track adoption metrics such as percent of AI suggestions accepted and override reasons, and report weekly KPI trends.