Why Enterprise AI fails (and what actually works) - Marc-Antoine Lacroix, Founder of Pivot

In this conversation, Marc-Antoine Lacroix, Founder of Pivot and former CTO and CPO at Qonto, explains why enterprise AI needs more than a powerful model: trusted data, the right skills, and a clear return on investment. He shares the moment an AI model invented missing transaction data instead of flagging a problem, and why that experience reinforced his conviction that reliable AI starts with the systems underneath it. We unpack Pivot’s approach to building an AI operating system for procurement, why Marc-Antoine believes an AI wrapper is not a moat, and how its AI Studio uses forward-deployed engineers to turn customer problems into agents with measurable value. 🚀 What you’ll learn: ⚡️ Why plugging AI into legacy systems is not enough to deliver value ⚡️ What happened when an AI model tried to reconcile inconsistent financial data ⚡️ The three layers behind Pivot: a system of record, a system of engagement, and AI agents ⚡️ Why better UX is essential to getting reliable data into your system ⚡️ How to train teams to check AI outputs instead of blindly trusting them ⚡️ Why Marc-Antoine believes interfaces still matter in an agentic world ⚡️ Where defensibility sits when competitors can rebuild your AI features ⚡️ How Claude and MCP changed the way Pivot customers analyse procurement data ⚡️ How Pivot’s AI Studio builds and tests custom agents with customers ⚡️ When custom work makes commercial sense for a SaaS company ⚡️ Why an API-first product is essential to the AI Studio’s autonomy ⚡️ What procurement teams should automate, and why key supplier negotiations still need humans 👉 Don’t miss the future of product leadership. Hit subscribe for more conversations with France’s boldest product thinkers. 📌 Follow me on LinkedIn for more insights and behind the scenes:   / cedblanchard   📌 Follow Marc-Antoine Lacroix:   / marc-antoinelacroix   📌 Learn more about Pivot: https://www.pivotapp.ai/ Chapters: 00:00 Enterprise AI: beyond the hype 00:48 Meet Marc-Antoine Lacroix, Founder of Pivot 01:33 Why enterprise AI projects fall short 02:59 The three layers of AI and a reconciliation gone wrong 07:31 Real value or expensive experiments? 09:09 Checking AI outputs and why interfaces still matter 15:30 Building the foundations alongside AI capabilities 16:26 Proving ROI before selling the hype 20:53 Procurement’s data problem starts with the user experience 24:02 Why an AI wrapper is not a moat 29:09 What changes when Claude can access trusted data 33:37 Inside Pivot’s AI Studio 34:54 Testing agents, proving ROI, and making custom work viable 36:55 Selling value, not people, and building API-first 38:56 What wins when the AI hype fades 41:14 Margins, valuations, and playing the long game 46:12 The real moat: trusted data and skilled people 49:24 Rapid fire: automation, suppliers, and human judgment 53:39 A French AI company to watch: Dust #ProductManagement #EnterpriseAI #AIAgents #Procurement #SaaS #Pivot