构建多智能体系统：金融服务领域的模式 [重复].pdf

《构建多智能体系统：金融服务领域的模式 [重复].pdf》由会员分享，可在线阅读，更多相关《构建多智能体系统：金融服务领域的模式 [重复].pdf（26页珍藏版）》请在三个皮匠报告上搜索。

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.I N D 3 7 4-RAlfredo CastilloMaira Ladeira TankeArchitecting multi-agent systems:Patterns in finan

2、cial servicesHe/HimSr Solutions Architect FSI GenAIAWSShe/HerTech Lead Agentic AIAWS 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AgendaWhy deploy multi-agent systems?Collaboration patternsWhiteboarding use case examplesHow this can help you 2025,Amazon Web Services,Inc.or its

3、affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Why deploy multi-agent systems?2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.What is an AI agent?Prompt ResultAgentModelToolsInvoke agentReturn final responseInvoke modelGet respons

4、e,reasoning,tool selection Execute toolReturn resultAgentic loop 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Single agents start small,but if you take it too farMicroservicedatabaseflooding policiesTerminologyThird-party serviceMachine LearningdatabaseUnderwriting AgentStart-a

5、pplicationget-documentationget-actuarial-datarules-validationUnderwriting ToolsCommission Calculation Toolsstart-new-commissionextend-commissionclose-commissionget-detailsAgency Broker Toolsenroll-agencyterminate-agencyRisk Modeling Toolsrisk-evaluationclaims-reserveChart-analysissubmit-reviewTermin

6、ology Knowledge BaseSearch TerminologyFlooding Knowledge BaseGet-flooding-policiesdatabaseCommission Service 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.it leads to challengesCoding gets complicated Complex prompts to limit hallucinations Fragile,hard to maintainAgent gets con