The first phase of AI in software engineering was simple: give developers AI coding tools and let experimentation spread.

The next phase is harder.

As coding agents take on larger tasks across files, systems and the whole lifecycle, work moves faster but coordination does not get simpler. Some developers become heavy users, others barely use the tools. Review and quality practices drift. New bottlenecks appear in architecture, context, testing, security and decision-making.

Leaders see activity rising without knowing whether outcomes improve. At this point, AI adoption becomes an operating-model challenge.

This session applies lessons from building an industrial agent factory to engineering organisations: keep each agent's task narrow enough to validate, give every agent a named owner, focus governance on what gets built and what is ready for release, and scale control with autonomy and risk. It also applies a lesson known since the 1980s: people cannot reliably take over a process they were never part of, which changes how review and supervision must be designed.

Participants get an AI-Native Engineering Leadership Framework to assess where they are today and what to change next.

Key learnings:

- Identify the organisational conditions required for agentic development to scale across teams
- Design review, supervision and ownership models for work shared by developers and agents
- Set governance and measurement priorities without creating bureaucracy