Ajith Suresh

Talk Title:

Building Enterprise AI That Organizations Actually Trust: Architecture, Explainability, and the Human Side of Intelligent Systems

Abstract:

The promise of enterprise AI is well understood. The gap between that promise and what organizations actually deploy and trust is less discussed but more consequential. This talk draws on the speaker’s published IEEE research and production experience building AI-driven analytics systems at global scale to address that gap directly.

The first part of the talk examines the architectural foundations of trustworthy enterprise AI, using the speaker’s work on LLM-powered conversational business intelligence as a primary case study. The system combining DistilBERT-based intent classification and entity extraction with a fine-tuned LLaMA-3-8B-Instruct model for SQL generation achieves 97.8% SQL execution accuracy and 94.1% multi-turn context retention on standard benchmarks, enabling non-technical business users to query complex enterprise data warehouses in natural language without SQL expertise or reliance on centralized data teams. The talk covers the architectural decisions that make this work at enterprise scale: unified pipeline design, semantic mapping, query validation, and conversational context retention across multi-turn interactions.

The second part addresses what the speaker identifies as the harder problem: organizational trust. A technically accurate system that people do not trust produces no decisions. Drawing on a production case study from Amazon’s Account Health Support organization where a fundamental measurement flaw in a legacy analytics system had gone undetected across six global sites until a ground-up rebuild corrected it the talk presents a framework for enterprise AI deployment that treats stakeholder trust-building, data governance, and explainability as first-class engineering deliverables alongside accuracy and throughput. The 90 percent reduction in missed calls and approximately $2 million in annualized cost savings that followed were not the result of engineering alone they were the result of engineering that the organization trusted enough to act on.

The talk closes with a forward-looking perspective on where enterprise AI is heading: the convergence of conversational interfaces, explainable AI frameworks, and responsible governance models that will define the next generation of intelligent enterprise systems. Attendees will leave with concrete architectural principles, a practical trust-building framework, and a clear-eyed view of the organizational conditions that determine whether AI investments produce decisions or just outputs.

Profile:

Ajith Suresh is a Data Analytics and AI Strategy professional building enterprise AI systems, LLM-powered business intelligence platforms, and operational analytics infrastructure across Amazon, Illumina, Dell Technologies, and McKesson.

 

At Amazon’s Account Health Support organization, he leads analytics for a global operation spanning six sites and over 860 specialists across North America, Europe, and Asia-Pacific. His work includes building the organization’s core performance measurement pipeline from scratch, designing LLM-powered reporting systems that generate executive-level insights automatically, and owning the Weekly Business Review framework relied upon by Senior Directors for operational decision-making. His analytics initiatives have contributed to a 90 percent reduction in missed calls and approximately $2 million in annualized cost savings.

 

His research on LLM-powered conversational business intelligence, explainable AI frameworks, and generative-AI-driven Auto-BI architectures has been published in IEEE conference proceedings. He is the author of Artificial Intelligence, Business Intelligence, and Analytics: Modern Approaches to Organizational Excellence (ScienceTech Xplore, 2026) and a Distinguished Fellow of the Soft Computing Research Society. A German Utility Patent on his AI-driven decision intelligence and LLM-based SQL orchestration architecture is currently in filing. He holds an M.S. in Business Analytics from Wichita State University and serves as a Lifetime Editorial Board Member of the International Journal of Artificial Intelligence, Data Science, and Machine Learning.