In this wide-ranging conversation, Dean Elwood, CEO of Umony, sits down with Alan Charbonneau, CTO of Umony, to explore one of the most pressing questions in today’s AI landscape: what does it really mean to trust intelligent systems?
From hallucinations and explainability to hybrid lexicons, human-in-the-loop workflows, and the limits of agentic systems, Dean and Alan break down where AI delivers, where it fails, and why progress may be shifting from “jobs” to “tasks.”
They discuss the plateau of model quality, the risks of synthetic data polluting the internet, the economics of failed AI initiatives, and whether we’re chasing AGI for innovation or for the trophy. Along the way, they examine how UX, curation, and “AI seasoning” may hold the key to making AI actually useful, safe, and trustworthy.
It’s a conversation about technology, but also about ethics, governance, and what remains fundamentally human as automation scales.
Chapters:
00:00 Intro
01:11 Trusting AI: Where Do We Begin?
03:35 Explainability, Citations & Transparency
06:44 Regulators, Risk & 100% Data Coverage
08:13 Beyond Red Flags: Business Insights & Green Flags
11:14 The Limitations of Lexicons & Fuzzy Models
13:44 The Future Without Lexicons
15:22 Human in the Loop: Why It’s Not Going Anywhere
20:09 AGI: A Goal or a Distraction?
23:39 Big Tech, Valuations & the Trophy Problem
25:59 Apple, Trust & Risk
30:36 The AI Hype Cycle & ROI Reality
33:05 Radiologists, Tasks & Human Judgment
34:57 Chat Interfaces vs Better UX
40:31 When AI Gets Things Wrong
44:58 The Future of Dashboards
47:29 AI in Small Doses