Date & time

17:00

You track your sleep, your spending, your workouts. Anything you want to improve, there’s an instrument for it — except the agent writing critical code, opening pull requests, and shipping features on your behalf. When working with AI agents, you’re still working off gut feel, with no real read on which models and skills are worth their cost, whether you’re getting faster, or how you compare to the developers who’ve already figured it out.

The days of unrestricted agent use are numbered as organizations introduce hard budgets per engineer. Longer-running agent sessions compound the problem: what used to be a one-line suggestion becomes a multi-step black box: planning, writing, testing, and opening the PR with no efficiency tracker.

This workshop puts that instrument in your hands. Working in a live GitKraken Insights sandbox loaded with real-world data, you’ll learn to read your own agentic workflow across adoption, autonomy, activity, cost, and outcomes. You will also be able to benchmark against your own organization.

Key takeaways:

• How to measure the adoption, autonomy, activity, cost, and outcomes of your own agentic workflow and know exactly what to improve next
• Which models and skills are worth their cost for different kinds of work, and where you’re likely overspending
• How to scale your agent work without scaling the bill