3321 - 从挑战到自动化解决方案 - 利用红帽人工智能释放智能体人工智能.pdf

1、1Unleashing Agentic AI with Red Hat AIFrom Challenges to Automated Solutions Philip HayesSenior AI Platform ArchitectYounes Ben BrahimRed Hat AI BU2Agentic AIAgentic AIUse CasesChatbots for FAQsFollows simple,predefined rulesRepetitive TasksLevel 0Retrieves data and recommends context-aware actionsL

2、evel 1Pulls from a knowledge base to recommend next steps for a support ticketInformation Retrievals AgentsLevel 2Schedules meetings and sends follow-up emails within a calendar/email systemAgents autonomously orchestrate low complexity tasks in a single data environment,can autonomously execute tas

3、ksTask Orchestration AgentsLevel 3Orchestrates workflows by connecting and harmonizing data across multiple business domains(ex.CRM or Finance)Manages a sales pipeline using harmonized data from CRM,service tickets,and financial reportsMulti-Domain OrchestrationLevel 4Specialized agents collaborate

4、to process orders,update inventory,and route customer feedbackA system of specialized agents collaborate in real-time to manage complex,cross-domain processesMulti-Agent Systems3From passive tools to active decision makers Agentic AI on Red Hat AI4AI agents integrate models,functions&toolsGen AI Mod

5、els,Predictive AI Models,Code Functions,Search&moreUserAI agentPromptResponseComplex taskNLP to understand prompt,generate task list and executeCodeExecution responseLLM12Code executor3Data,document or webAgentic AI on Red Hat AIAgent=Model(Powerful LLM(s)+Augmentation(Tools,Memory,RAG)+Loop(A conti

6、nuous process of thinking and acting)5Lets Pasta!Need Ingredients.Oops,I might need to look it upGot it!Lets cook,I need to use a pan,pot,whisk!Lets Cook!TastingI am lovin it!Nah!6GoalTaking Action:Searching knowledge(Can be tools)Taking Action:Using ToolsObserving/PerceivingResultsNah!Thinking7Appl