Tuesday, February 09, 2027
9:00-9:45am
Tuesday - 2/09/27
|
|
10:00-10:45am
Tuesday - 2/09/27
|
AI Agent Event (Room 250C)
Permission-Aware Enterprise AI Agents: Designing Governed Autonomy for Production
AIA-02 Giving agents autonomous tool access introduces severe permission risks. This session details architectural patterns for building identity-aware, permission-aware agents that operate across corporate APIs, knowledge repositories, and issue trackers without bypassing enterprise access controls. Learn to implement decision boundaries, human-in-the-loop checkpoints, workflow tracing, and safe fallbacks.
Register me 
|
11:00-11:45am
Tuesday - 2/09/27
|
AI Agent Event (Room 250C)
From Assistants to Systems: Designing Reliable AI Agents for Real-World Engineering Workflows
AIA-03 Single-prompt agents struggle with consistency on complex tasks. This session presents design patterns for building reliable agentic systems by breaking workflows into staged phases: planning, execution, validation, and review. Learn how to implement multi-agent orchestration, schema-driven validation, and context-aware memory for automated code analysis, refactoring, and reviews.
Register me
|
12:00-12:30pm
Tuesday - 2/09/27
|
|
1:30-2:15pm
Tuesday - 2/09/27
|
AI Agent Event (Room 250C)
Why AI Agents Fail Before They Reason: A Self-Curating Multi-Agent Framework for Reliable Enterprise AI
AIA-05 Agent failures often stem from bad context rather than flawed reasoning. This session introduces a closed-loop multi-agent curation framework featuring Architect, Extractor, Critic, and Orchestrator roles. Learn how distilling frontier teacher models into specialized student models slashes context volume by 17× and speeds up processing—achieving sub-20ms execution times while keeping guardrails debuggable.
Register me
|
2:30-3:15pm
Tuesday - 2/09/27
|
AI Agent Event (Room 250C)
Moving agentic systems from prototype to production requires moving past untraceable loops and brittle exception handling. This talk focuses on building accountable workflows using deterministic controls, observable agent handoffs, and explicit human escalation checkpoints—providing a practical engineering framework to ship governed, production-grade agents.
Register me
|
3:30-4:00pm
Tuesday - 2/09/27
|
|
Wednesday, February 10, 2027
9:00-9:45am
Wednesday - 2/10/27
|
|
10:00-10:45am
Wednesday - 2/10/27
|
AI Agent Event (Room 250C)
From Operators to Orchestrators: A Two-Stage Framework for Deploying Agentic AI in Enterprise Operations
AIA-09 Over 40% of agentic projects stall due to brittle workflows and lack of operational oversight. Grounded in a Fortune 500 deployment, this session introduces a two-stage operational framework: a 5-dimension readiness diagnostic to evaluate workflows before model selection, and an agentic surface design that positions human intervention precisely where agents are weakest.
Register me
|
1:30-2:15pm
Wednesday - 2/10/27
|
AI Agent Event (Room 250C)
Unmanaged agent context windows degrade reasoning quality and waste tokens. This session explores Context Engineering—moving from passive prompt pushing to a stateful, demand-driven memory model. Core pillars include: wrapping LLMs in stateful harnesses, using Model Context Protocol (MCP) for Just-In-Time tool injection, and navigating compressed histories using pointer schemas to fetch raw source data on demand.
Register me
|
2:30-3:15pm
Wednesday - 2/10/27
|
AI Agent Event (Room 250C)
Beyond AI Assistants: Building Production-Grade Agentic AI for Autonomous Enterprise Decisions
AIA-11 Transitioning from basic RAG chatbots to enterprise-scale autonomous systems introduces complex architectural challenges. This session covers core engineering patterns required to deploy production-grade agents: multi-agent decision workflows, dynamic context management, tool integration via APIs, memory and reasoning modules, and cloud-native observability frameworks.
Register me
|
Thursday, February 11, 2027
9:00-9:45am
Thursday - 2/11/27
|
|
10:00-10:45am
Thursday - 2/11/27
|
AI Agent Event (Room 250C)
High-stakes operations cannot rely on unconstrained autonomous reasoning. This practical case study highlights an architecture running six specialized investigation agents across 400+ dataset types. Discover why capping agents at two reasoning steps prevents hallucination loops, how deterministic contradiction rules catch agent disputes without secondary models, and how to implement template fallbacks.
Register me
|
11:00-11:45am
Thursday - 2/11/27
|
AI Agent Event (Room 250C)
As AI transitions from static response generation to autonomous tool invocation and multi-agent execution, traditional IAM and application security fall short. This session examines dynamic runtime security threats—including prompt injection, compromised context, excessive tool permissions, and poisoned data feeds—and delivers practical safeguard strategies to enforce secure agent action in production.
Register me
|
