Watch the podcast recording below

In this episode of the Discussing New Approaches podcast, Josh Cozens from WeBuild-AI joins Eric Walker-Rawson to talk about what it takes to ship production-grade AI in highly-regulated enterprises, drawing on real client work, not theory.

You can now build in three months what used to take six to nine, but if your governance process, company structure, and ideation discipline haven't evolved to match that pace, the build speed doesn't matter.

Josh shares why 120 governance tickets spread over six to nine months is manageable, and solved by addressing as an organisation.

He also covers ruthless ideation: how one client went from 50 ideas down to three that actually made it to production, and why that discipline is what makes delivery possible.

Josh also covers spec-driven development, build vs buy decisions, and guardrailing agentic workflows: the choices that determine whether AI ships or stalls. All these choices build into deliberate AI - a method that means custom, compliant and production-grade AI is guarunteed.

Whether you're leading an AI programme, stuck between demo and production, or trying to align innovation days with what teams can actually use, this conversation offers a practical view of AI-native delivery at enterprise pace.

About the speaker:

Joshua is a Principal Consultant at WeBuild-AI.

Joshua helps regulated enterprises design and deliver production-grade AI, from platform builds on secure cloud infrastructure to getting adoption right in the real world.

His recent work includes taking a client from disparate spreadsheets to a secure, scalable AI platform in months, not years, and treating organisational blockers, adoption dips, and specification quality as core delivery problems, not afterthoughts.