The word "agentic" is everywhere right now. Vendors are slapping it on everything. But when you strip the pitch away, what does an AI agent actually do inside an enterprise — and why does it keep breaking when someone tries to put it into production?

In this episode, Carl Andersson walks through what agentic AI really means in operational terms: how agents differ from simple automations, what makes them genuinely useful, and what makes them genuinely dangerous in environments where accuracy and auditability matter.

We look at where agentic AI is already working — in finance, legal, and insurance workflows — and trace the failure modes that trip up most deployments before they reach the team that was supposed to benefit.

IN THIS EPISODE

• What separates an AI agent from a workflow automation
• Why agentic AI fails most often at the boundary between systems, not inside them
• The auditability problem in regulated industries
• What "human in the loop" actually means in practice — and when it becomes a liability
• How to scope your first agentic deployment without over-engineering it

LINKS

• Nutropic AI

NUTROPIC

Nutropic is a Melbourne-based AI consultancy founded by Carl Andersson, a technologist with 25 years across software engineering, business technology and process design, primarily with Australian finance and insurance organisations. The practice helps teams, workflows and technical systems adopt AI in ways that produce lasting outcomes rather than impressive-looking activity.

LINKS

• Nutropic AI
• LinkedIn
• Twitter
• Substack
• YouTube
• Threads
• Bluesky
• Instagram

LISTEN

• Spotify
• YouTube
• Apple Podcasts
• Amazon Music
• Podcast Addict
• Pocket Casts
• Player FM