Agentic AI is rapidly moving into production across most enterprises. Teams are building agents, deploying co-pilots, and exploring autonomous workflows at an accelerating pace. Yet for many, the real challenge is only beginning.

The question is no longer “What can we build?” It is “What actually works at scale?”

Despite significant investment, most enterprise AI initiatives are struggling to move from promising pilots to reliable, production-grade systems. Engineers and architects are grappling with fragmented data environments, immature orchestration layers, rising infrastructure costs, and the growing complexity of managing multi-agent systems. At the same time, senior-level technology executives are being asked to ensure governance, security, and measurable business impact - often without clear standards or proven playbooks.

In this highly interactive, member-driven working session, the event moderators will facilitate a structured, open exchange where all attendees are expected to actively contribute to the group conversation - sharing what is working, what is not, and to engage in a candid, group problem-solving dialogue around the most pressing challenges facing enterprise AI initiatives.

Please Note - this is not a passive event. Throughout the event, all attendees are expected to actively participate in the discussion and contribute real-world challenges and solutions.

Topics of discussion will include, yet will not be limited to:

• Moving from pilots to production: what is breaking at scale
• Orchestrating multi-agent systems and managing failure modes
• Data readiness, trust, and the role of the semantic layer
• Cost, infrastructure constraints, and model optimization
• Security, governance, and controlling agent sprawl
• Closing the gap between AI capability and organizational readiness

Rahul Chawla, Lead Business Intelligence Data Warehouse Engineer, GOOGLE

Prasad Maderamitla, Senior & Principal Quality Engineering Leader, Artificial Intelligence, SALESFORCE

Ashok Pal Singh Rathore, Senior Technical Program Manager, FACEBOOK (META)

Robin Patra, Head Data & Analytics Manager, KEELEY COMPANIES

LLM and GenAI capabilities are rapidly being embedded into enterprise workflows, customer-facing applications, and internal decisioning systems. As adoption accelerates, organizations are encountering a new class of security, governance, and control challenges that extend far beyond traditional application or data protection models. Systems that appear functional in early deployment stages are increasingly exposing risks around data leakage, prompt injection, uncontrolled outputs, and inconsistent policy enforcement when scaled across real-world environments.

The challenge is no longer simply enabling AI - it’s controlling how these systems behave in production. This includes defining where guardrails must be enforced, how sensitive data is accessed and protected, how outputs are validated, and how governance models evolve alongside rapidly changing AI capabilities. At the same time, business and technology leaders must align on acceptable risk, operational accountability, and measurable impact - often without established standards or fully mature tooling.

In this highly interactive, member-driven working session, the event moderators will facilitate a structured, open exchange where all attendees are expected to actively contribute to the group conversation - sharing what is working, what is not, and to engage in a candid, group problem-solving dialogue around the most pressing challenges facing enterprise LLM & GenAI security, guardrails, and governance at scale.

Please Note - this is not a passive event. Throughout the event, all attendees are expected to actively participate in the discussion and contribute real-world challenges and solutions.

Topics of discussion will include, yet will not be limited to:

• Where LLM and GenAI systems are breaking down from a security and control perspective in production environments, including prompt injection, data leakage, and uncontrolled outputs
• How organizations are implementing guardrails at different layers of the stack, including input validation, output filtering, access controls, and policy enforcement
• The challenges of governing AI systems across business and technical functions, including ownership, accountability, and risk alignment
• How sensitive enterprise data is being accessed, restricted, and monitored within GenAI workflows
• The approaches being used to balance security, usability, and performance without over-constraining system effectiveness
• How organizations are operationalizing governance frameworks in environments where standards, tooling, and best practices are still evolving

Sundeep Bobba, Cloud DevOps & Enterprise Systems Engineering Leader, SOUTHWEST AIRLINES

Malavika Lakireddy, Global Director of Product Management, UBER

Sibasis Padhi, Staff Software Engineer, Walmart Global Tech, WALMART

Navita Singh, Principal Data Scientist - Autodesk Research, AUTODESK

Early-stage Agentic AI systems are delivering impressive initial results, but many are breaking down when exposed to real enterprise environments. What appears to work in controlled settings quickly becomes unstable when agents are required to interact with fragmented data systems, invoke multiple tools, and operate across longer, multi-step workflows. Outputs become inconsistent, failure points multiply, and debugging becomes increasingly opaque- leaving engineering and architecture teams stuck between promising prototypes and unreliable production systems.

In response, organizations are beginning to formalize how agent-based systems are designed, structured, and governed - moving beyond prompt-level experimentation toward full architectural frameworks that account for orchestration, memory, tool usage, and failure handling. This includes rethinking how agents manage state, how decisions are evaluated, and how systems are instrumented for observability and control, enabling more predictable, scalable, and production-ready implementations.

Please Note - this is not a passive event. Throughout - and at the end of the speakers’ presentation, all attendees are expected to actively participate in the discussion and contribute real-world challenges and solutions.

Topics of discussion will include, yet will not be limited to:

• Designing multi-agent systems to handle complex workflows, including orchestration logic, tool chaining, and state management across tasks
• Identifying where agent-based systems break down in production environments, including failure modes such as hallucinations, tool misuse, latency issues, and inconsistent outputs - and implementing mitigation approaches
• Integrating enterprise data systems into agent workflows, including retrieval strategies, API interaction patterns, and data access constraints
• Implementing guardrails, observability, and evaluation frameworks to improve reliability, traceability, and system performance over time

Santhosh Vijayabaskar, Director, Intelligent Automation & Process Excellence, FIDELITY INVESTMENTS

The mandate sounds simple: move faster, deliver more, and do it with fewer people. But inside enterprise AI initiatives, that mandate is colliding with a very different reality. Teams are being asked to operationalize Agentic AI, rebuild data foundations, and integrate entirely new capabilities into legacy environments - while existing delivery models quietly break under the weight of it all. What used to be predictable roadmaps have become moving targets, and even the most experienced leaders are finding that traditional resourcing, planning, and execution strategies no longer hold.

In response, leading organizations are starting to rethink how work actually gets done - shifting toward leaner, more adaptive team structures, redefining roles across engineering and product, and prioritizing fewer, higher-impact initiatives that can realistically be delivered. In turn, leaders are navigating these constraints in real time, making trade-offs across priorities, and recalibrating execution models to maintain momentum without overextending teams or compromising outcomes.

Please Note - this is not a passive event. At the end of the speakers’ presentation, all attendees are expected to actively participate in the discussion and contribute real-world challenges and solutions.

Topics of discussion will include, yet will not be limited to:

• How enterprise teams are restructuring roles and responsibilities to support AI-driven initiatives without expanding headcount
• Where traditional project management and delivery models are failing - and how leaders are adapting execution frameworks in response
• How organizations are prioritizing AI and data initiatives under constrained resources while maintaining measurable business impact
• The trade-offs leaders are making between speed, quality, and scalability - and how those decisions are playing out in production environments

Bobbie Long, Senior Vice President, Global Engineering Lead, Digital Automation & Infrastructure Services, Quality, CITI

Ashok Pal Singh Rathore, Senior Technical Program Manager, FACEBOOK (META)

Suzanne LaLena, Senior Vice President, Automation Engineering, BNY

Elezebeth Daisy George Daniel, Service Delivery Leader, Supply Chain, SOLENIS

Moderator: Gaurav Basra, Chief Executive Officer, BASRA CONSULTING SERVICES

Enterprise Retrieval-Augmented Generation (RAG) systems are being deployed rapidly, but many are failing to deliver consistent, reliable outputs when exposed to real-world data environments. What begins as a straightforward retrieval pipeline often becomes unstable as data quality issues, poor chunking strategies, weak embedding alignment, and inefficient indexing approaches compound. As system complexity increases, retrieval precision degrades, latency rises, and outputs become less grounded - creating significant challenges for engineering and architecture teams responsible for production performance.

As these challenges surface, organizations are moving toward more rigorous approaches to designing, tuning, and validating RAG systems - placing greater emphasis on data preprocessing pipelines, retrieval optimization, indexing strategies, and evaluation methodologies. This includes refining how data is chunked and embedded, how vector search is tuned, and how retrieval outputs are validated and monitored, enabling more deterministic, scalable, and high-performing systems in production environments.

Please Note - this is not a passive event. At the end of the speakers’ presentation, all attendees are expected to actively participate in the discussion and contribute real-world challenges and solutions.

Topics of discussion will include, yet will not be limited to:

• How data ingestion and preprocessing pipelines are being structured, including chunking strategies, metadata enrichment, and embedding model selection
• Where retrieval pipelines break down at scale, including issues with vector similarity, context dilution, and irrelevant document surfacing - and how these are being addressed
• How indexing and vector database strategies are being optimized for performance, including hybrid search approaches, re-ranking techniques, and latency considerations
• The evaluation and monitoring frameworks being implemented to measure retrieval quality, grounding accuracy, and system performance over time

Chirag Shah, Global Chief Information Security Officer & Data Protection Officer, MODEL N

Deep Patel, Senior Data Engineering Lead, ROBINHOOD

Gladson Baby, Vice President & Director of Data & AI Enablement, Intelligent Automation & System Integration, FIFTH THIRD BANK

Pradeep Madhavankutty, Director, AI Delivery & Product Development, STANDARD CHARTERED

Moderator: Jacqueline Starr, IT Business Analysis Senior Specialist, CNA INSURANCE

Over the past several years, enterprises have made significant investments in data platforms, governance frameworks, and analytics capabilities. Yet as AI initiatives accelerate, many organizations are finding that their data environments remain fragmented, misaligned, and difficult to operationalize at scale. What was intended to be a strategic asset is often still functioning as a constraint - slowing down AI adoption, limiting visibility, and creating disconnects between technical capabilities and business priorities.

As AI becomes more deeply embedded into core business functions, data strategy is being reframed as a business-critical discipline rather than a technical function. Organizations are redefining how data is governed, prioritized, and funded - placing greater emphasis on aligning data initiatives to revenue generation, risk management, and operational performance. This shift requires clearer ownership models, tighter integration between business and technology teams, and more disciplined approaches to measuring the impact of data investments across the enterprise.

Please Note - this is not a passive event. At the end of the speakers’ presentation, all attendees are expected to actively participate in the discussion and contribute real-world challenges and solutions.

Topics of discussion will include, but not be limited to:

• How enterprise data strategies are being realigned to directly support AI initiatives, revenue growth, and operational performance
• Where existing data investments are failing to deliver business value, and how organizations are reassessing priorities, funding models, and execution approaches
• How ownership and accountability for data are being structured across business and technology functions to improve alignment and outcomes
• The approaches being used to measure the business impact of data initiatives, including performance metrics tied to revenue, risk, and efficiency

Amit Shivpuja, Director of Data Product & Artificial Intelligence Enablement, WALMART

Sundar Krishnan, Director, Artificial Intelligence & Machine Learning, OPTIMUM

Sai Krishnan Mohan, Vice President, Data & Analytics, BAJAJ AUTO

Subba Rao Katragadda, Senior Principal Data Engineer, JOHNSON & JOHNSON MEDTECH

Moderator: Vijay Jakka, Former Digital, AI, Analytics & Data Governance Global Leader, METLIFE