Have we reached the point where AI can genuinely build software without developers writing code?

In this episode of The Common Sense CTO Podcast, Matt Stephenson sits down with digital transformation and AI specialist Mark Strefford to explore a question that is rapidly moving from theory to practice: what happens when software development becomes an exercise in specification, validation and orchestration rather than programming?

Matt has been openly sceptical about claims that AI will replace experienced engineers. Mark, meanwhile, has spent the last two years building products almost entirely through AI-driven workflows. Their conversation reveals just how quickly the landscape is changing.

Together they discuss:

• Why AI coding has progressed dramatically in the last six months
• Whether developers are becoming designers, reviewers and orchestrators rather than coders
• The emergence of AI agents acting as members of a software delivery team
• Building AI-native software delivery lifecycles with automated governance and review processes
• Accountability, quality assurance and human oversight in AI-generated systems
• Why acceptance criteria and engineering discipline may matter more than ever
• The economics of AI-assisted development and the reality of token-based pricing
• Local models, edge AI and whether organisations will eventually run their own AI infrastructure
• The uncomfortable question of who is accountable when AI-generated software fails

Mark shares practical examples from his own workflow, including autonomous coding agents, AI-driven testing, multi-agent review processes and the surprising extent to which he now judges software by outcomes rather than source code.

For Matt, the conversation represents a shift in perspective. While many of the concerns around quality, governance and accountability remain, it is becoming increasingly difficult to dismiss the possibility that AI-assisted engineering is evolving into something much bigger.

If you're a CTO, engineering leader, architect, product professional or experienced developer trying to separate hype from reality, this episode offers an honest and nuanced discussion about where software engineering may be heading next.

Perhaps the biggest question isn't whether AI can write code.

It's whether our delivery processes, governance models and organisational structures are ready for what comes after that.