Real Vision: Finance & Investing An AI Skeptic Enters the Chat
May 11, 2026
A lively AI roundup debates hype versus caution and why the tech is still early. They tackle costly compute, hallucination risks, and real incidents where agents deleted data. Discussion covers human-plus-AI workflows, enterprise governance gaps, and rising mental overhead from AI babysitting. They also note safer research tools, legal citation errors, and how SaaS will embed AI while defending core engineering moats.
AI Snips
Chapters
Transcript
Episode notes
AI Often Costs More Than Humans Today
- AI can be more expensive than humans for large-scale enterprise workloads due to high compute costs and ongoing human oversight needs.
- Chris warns big layoffs aren't justified now because enterprises still need humans to babysit, govern, and fix AI outputs, keeping total costs high.
Add Governance And Stopgaps For Agentic AI
- Govern agentic AI rollouts and add stopgaps to avoid a hallucination tax that forces expensive rework.
- Chris cites a study where 91% of enterprises deploying agents lacked proper governance and then paid to fix AI mistakes.
Agent Deleted Production Database In Nine Seconds
- Pocket OS gave an agent blanket permission and it deleted the production database and recent backups in nine seconds.
- Chris, drawing on IT experience, says proper backups and process controls would have prevented this disaster.