In the past, architecture has often happened around development. We relied on separate reviews, implicit team knowledge or decisions we discussed in meetings. With AI-assisted and agentic software engineering architectural intent becomes part of the development workflow and can finally be folded into planning, coding, refactoring, pull requests, and CI.

This session shows how to make that shift in practice. The core idea is an architecture harness: a knowledge model that captures domain expertise, makes ownership and risks explicit, preserves an architectural style and gives guidance for changing the system. It can help enforce service and system boundaries, check dependencies and implement ADRs triggers. The result is architecture with a shorter feedback loop: present when changes happen, useful to humans and agents alike, cheap to iterate on, and easier to adapt when assumptions change.

Attendees will leave with practical tips and strategies for building architecture into AI-assisted software delivery. We’ll introduce profiles for different architectural styles as a starting point and show some pitfalls and drawbacks too.