The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch 20VC: Enterprises Will Not Adopt AI without Forward-Deployed Engineers | Who Wins the Data Labelling Race: How Does it Shake Out? | How Synthetic Data Threatens the Future of Human-Generated Data with Matt Fitzpatrick, CEO of Invisible Technologies
513 snips
Dec 31, 2025 
Matt Fitzpatrick, CEO of Invisible Technologies and former McKinsey senior partner, dives into the crucial role of forward-deployed engineers in AI adoption. He argues that enterprises can’t fully embrace AI without these experts, highlighting the gap between AI model advancements and slow enterprise deployment. The conversation also covers the evolving data labeling landscape, the importance of quality data, and how human feedback remains invaluable in AI training. Fitzpatrick's insights emphasize the need for specialized strategies as AI reshapes industries.
AI Snips
Chapters
Books
Transcript
Episode notes
25M Returns Agent That Hallucinated
- An e-commerce retailer spent $25M building a returns agent and measured success by speed and sentiment.
- The agent later hallucinated costly outcomes and the company reverted to deterministic flows.
Price GenAI Against User Acceptance
- Accept that GenAI projects resemble ML deployments: build, validate, and charge upon user acceptance.
- Structure contracts so payment occurs after UAT and proven accuracy, not upfront subscription fees.
Training And Fine-Tuning Use One Assembly Line
- AI training and enterprise fine-tuning share the same digital assembly-line processes for sourcing experts, validating outputs, and scaling.
- Invisible leverages its 1.3M expert marketplace to serve both model builders and enterprise fine-tuning needs.
