Use different agents to validate AI code

What really happens when AI helps teams write code faster, but everything else in the delivery process starts to slow down? In this episode of Tech Talks Daily, I'm joined once again by returning guest and friend of the show, Martin Reynolds, Field CTO at Harness.

It has been two years since we last spoke, and a lot has changed since then. Martin has relocated from London to North Carolina, gaining back hours of his working week. Still, the bigger shift has been in how AI is reshaping software delivery inside modern enterprises.

Our conversation centers on what Martin calls the AI velocity paradox. Development teams are producing more code at speed, often thanks to AI coding agents, yet testing, security, governance, and release processes are struggling to keep up. The result is a growing gap between how fast software is written and how safely it can be delivered.

Martin shares research showing how this imbalance is already leading to production incidents, hidden vulnerabilities, and mounting technical debt.

We also dig into why this AI-driven transition feels different from previous waves, such as cloud, mobile, or DevOps. Many of the same concerns around security, trust, and control still exist, but this time, everything is happening far faster. Martin explains why AI works best as a human amplifier, strengthening good engineering practices while exposing weak ones sooner than ever before.

A significant theme in the episode is visibility. From shadow AI usage to expanding attack surfaces, Martin outlines why security teams are finding it harder to see where AI is being used and how data is flowing through systems. Rather than slowing teams down, he argues that the answer lies in embedding governance directly into delivery pipelines, making security automatic rather than an afterthought.

We also explore the rise of agentic AI in testing, quality assurance, and security, where specialized agents act like virtual teammates. When well-designed, these agents help developers stay focused while improving reliability and resilience throughout the lifecycle.

If you are responsible for engineering, platform, or security teams, this episode offers a grounded look at how to balance speed with responsibility in an AI-native world. As AI becomes part of every stage of software delivery, are your processes designed to safely absorb that change, or are they quietly becoming the bottleneck?

Useful Links

• Learn More About Harness
• The State of AI in Engineering
• The State of AI Application Security
• EngineeringX
• Follow Harness on LinkedIn
• Connect With Martin Reynolds

Thanks to our sponsors, Alcor, for supporting the show.