Parul Bajpai: Lessons from Building an AI Chatbot

Session Overview

"The real transformation isn't that AI can talk naturally. It's that AI can finally solve real customer problems."

In 2019, Parul's team started with a simple bot that could only handle basic intent recognition. Fast forward to 2025, and they have built a global AI ecosystem handling millions of conversations across 190+ countries. In this candid session, Parul Bajpai shares the journey of evolving from "Intent-based NLP" to "Agentic AI."

Parul discusses the "NLP Ceiling"—the point where traditional chatbots fail to scale—and how Generative AI unlocked the ability to handle multi-turn reasoning and complex problem resolution. She outlines the operational realities of deploying AI at an enterprise scale, emphasizing that "Guardrails are more important than the Model" and sharing her "Start Small, Scale Fast" rollout strategy.

Key Takeaways

03:15 The NLP Ceiling: Why intent-based bots break at scale (couldn't handle ambiguity or multi-turn context).
09:40 GenAI Transformation: The shift to the 3 pillars of modern bots: Natural Conversations, Knowledge Grounding, and Agentic Workflows.
15:20 Operational Excellence: The strategy of rolling out AI to 5% of users first, monitoring failure modes, and then scaling to 100%.
21:00 Safety First: Why guardrails, context blocks, and high-risk detection are critical for enterprise-ready AI.
27:30 Future Outlook: Moving from "Chatbots" to "Virtual Advisers" that proactively solve problems before the user even asks.