1711 - 智能制造 - 无人值守工厂的数据层.pdf

1、Orlando,FLOctober 69IBM TechXchange 2025How a Edge Data Fabric Enables Agentic AI for ManufacturingThe“BRAIN”of Agentic AISession code 1711Agentic Manufacturing-a Data Layer for Lights-Out FactoriesMark DavidsonAnylog,VP ProductsPresentation Presentation OverviewOverviewUnderstanding Agentic AI and

2、its role at the edgeFuture data trends in Agentic AICognitive processes driving industrial decisionsNeuralEdge Data Fabric as the AI brainFederated learning and multi-agent decision makingReal-world applications and ConclusionsWhat“Agentic AI”MeansWhat“Agentic AI”MeansAgentic AIAgentic AI refers to

3、AI systems that act as agentsagents:Observe Observe Reason/PlanReason/Plan ActActUnlike traditional LLM models that just respond to prompts,agentic AI systems have:MemoryMemory GoalsGoals Tools/actuatorsTools/actuatorsThink of them as decisiondecision-making making digital workersdigital workers rat

4、her than just LLM-based“chatbotsAt the EdgeAt the EdgeWhen we say at the edge,we mean running on edge devices or local running on edge devices or local gatewaysgateways instead of relying on cloud servers.Processing is close to where data is generated:IoT sensors,PLCs,industrial controllers,cameras,

5、or factory-floor gateways.Benefits:Benefits:Low latencyLow latency ResilienceResilience Privacy/securityPrivacy/security EfficiencyEfficiencyWhat is Agentic AI at the Edge?Inferencing at the edgeGPU at the edge needed for inferencingLocal Data stores to fulfill inferencing needsCloud access to edge

6、data for trainingLLMs-SLMsIntegration of LLM/SLM,reinforced learning and MLFederated Learning for edge modelsFewer data“round trips”to the cloudData Sovereignty and PrivacyData grows forever!BenefitsBenefitsCompetitive AdvantageImprove Business outcomesLower costsFuture Data Trends with Agentic AIAd