72x Faster Software Delivery with a Former AI Skeptic

On This Podcast

In this conversation, we examine Odevo's decision to rebuild five years of legacy property management software from the ground up, aiming to deliver eight years' worth of planned development in just eleven months with a third of the original headcount. Dominic explains that the ageing codebase had grown organically complex, and that a fresh, industry-standard tech stack designed specifically for agentic engineering offered a far better foundation than trying to bolt AI onto an inconsistent legacy system. We also discuss how the team scoped the project in two phases, targeting feature parity with the old product within five months and full three-year functionality by December 2026. Dominic is candid that a rewrite of this ambition would have seemed laughable to him just months earlier, underlining how quickly the calculus around large rewrites has shifted.

In this conversation, we examine how Odevo kicked off the rebuild with roughly two weeks of intensive AI-driven analysis across fifteen legacy projects, extracting business logic, data schemas, and entity relationships rather than the underlying code itself, since the new system uses an entirely different tech stack. We also discuss the follow-on discovery process, where every planning meeting was recorded and transcribed, then fed into a documentation system that automatically linked discussion points to specs and epics, meaning no one needed to type up notes by hand. Dominic describes this as one of his standout realisations: even chaotic, cross-cutting conversations could be untangled by capable models into clean, structured documentation with barely any manual editing. The approach front-loads effort into getting foundational decisions, such as roles, permissions, and data hierarchy, right early, since mistakes there are costly to unwind months later.

In this conversation, we examine the personal toll and eventual reward of Dominic's transition to a role where he has not written a single line of code by hand since January. He describes going through what he calls an ego death, grieving the loss of an identity built around writing elegant syntax, and openly admitting he doubted whether he even enjoyed the new way of working during the early months. We also discuss how that feeling has shifted over time, with Dominic now finding satisfaction in different rewards, such as rapidly iterating on prototypes with a designer or delivering a fully refined epic, rather than the dopamine of solving a technical puzzle by hand. Both speakers reflect on how this trade-off, delivering large, well-guarded chunks of working software instead of individually crafted code, represents a genuinely different, though still meaningful, form of engineering satisfaction.