<?xml version="1.0" encoding="utf-8"?>
<oembed>
  <version>1</version>
  <type>rich</type>
  <provider_name>Libsyn</provider_name>
  <provider_url>https://www.libsyn.com</provider_url>
  <height>90</height>
  <width>600</width>
  <title>Why AI Agents Fail at CPQ — and How to Fix It with Dr. Sanjay Mittal</title>
  <description>Can you trust an AI agent to configure your products, apply your pricing rules, and generate an accurate quote — every single time? According to Dr. Sanjay Mittal, founder of Predictika.ai and one of the original pioneers of constraint-based product configuration, the answer right now is: not without guardrails.  In this episode of the CPQ Podcast, Sanjay breaks down one of the most underappreciated risks in enterprise AI adoption — logic hallucinations. Unlike factual hallucinations, where an AI simply makes up information it doesn't have, logic hallucinations happen when an AI has all the right facts but still applies your business rules, constraints, and pricing policies incorrectly. For CPQ, where 100% correctness and completeness are non-negotiable, that gap can mean revenue leakage, incompatible configurations, and costly disputes with customers.  Sanjay speaks from deep experience. He founded Selectica in 1996, building constraint-based CPQ solutions for Cisco, IBM, GE, and Dell — guaranteeing 100% configuration correctness at scale. Cisco ran on their platform for 15 years. Today, at Predictica.ai, he's applying the same deterministic reasoning engine as a co-pilot alongside large language models, letting LLMs handle natural conversation while a constraint engine validates every output before it reaches the customer.  In this conversation we cover:  The difference between factual and logic hallucinations in LLMs Why probabilistic AI and deterministic CPQ are fundamentally mismatched What &amp;quot;correct and complete&amp;quot; really means in product configuration How a deterministic co-pilot architecture bridges that gap What questions enterprise buyers should ask any CPQ vendor adding AI   If you're evaluating AI-powered CPQ, building one, or just trying to understand where the hype ends and the real engineering begins — this episode is for you.  🔗&amp;amp;nbsp;Learn more about predictika at&amp;amp;nbsp;https://predictika.com/&amp;amp;nbsp;  📬&amp;amp;nbsp;Contact Sanjay&amp;amp;nbsp;on LinkedIn at&amp;amp;nbsp;https://www.linkedin.com/in/sanjaymittal/&amp;amp;nbsp; </description>
  <author_name>CPQ Podcast</author_name>
  <author_url>http://www.novuscpq.com</author_url>
  <html>&lt;iframe title="Libsyn Player" style="border: none" src="//html5-player.libsyn.com/embed/episode/id/41827725/height/90/theme/custom/thumbnail/yes/direction/forward/render-playlist/no/custom-color/88AA3C/" height="90" width="600" scrolling="no"  allowfullscreen webkitallowfullscreen mozallowfullscreen oallowfullscreen msallowfullscreen&gt;&lt;/iframe&gt;</html>
  <thumbnail_url>https://assets.libsyn.com/secure/content/203402565</thumbnail_url>
</oembed>
