Agentic Coding In Practice

Agentic Coding in Practice

AI coding tools have taken a major leap forward over the past months and teams are experimenting with new software engineering practices. We are facing the question “How do we create workflow and tooling around AI so the result is reliable and appropriately safe while optimizing use of human time?”

Many of us feel uneasy about whether we are making the right choices. We also lack visibility into what others are really doing, and what they stopped doing because it didn’t work. Let’s fix that!

What We’ll Cover

This session will go deep into the practical mechanics of agentic coding, from spec to PR to QA. Presenters will demo how they are using agents, skills, rules, hooks, instruction files, code review loops, test workflows, MCP tooling, worktrees, persistent context, and checkpoints to make AI coding tools work inside real engineering processes.

Where Human Effort Pays Off

We’ll focus on where people should and should not spend their time. As agents generate more (or all) of the code, humans need to focus on the leverage points: defining clear acceptance criteria, shaping implementation plans, enforcing architecture, identifying risk, improving test coverage, reviewing high-impact changes, and “sharpening the axe” – refining the agentic process itself.

Topics will include:

How presenters’ teams are working today: what tooling, workflows, and techniques are they using to deliver..
Where human review belongs in an agentic coding workflow
What quality gates teams rely on, including code review, testing, and red-green workflows
What teams adopt off-the-shelf, what they build themselves, what they stopped using, and how they improve their approaches over time.
How teams co-author specs, acceptance criteria, and implementation plans with agents
How engineers manage concurrent AI coding sessions

This will be a great session for engineering, product, and QA leaders who will leave with a clearer picture of what mature agentic coding looks like, and where your team has the most to gain.