AI-Native Software Delivery

Chapter 1. The Road to AI-Native DevOps

Most software development teams can tell war stories about deployments gone wrong. These are the stories that put us on paths to modernize our delivery practices.

Here’s one: After weeks or months of feature work, extensive refactoring, and a long testing and stabilization phase, a team is ready to deploy. Developers, operations team members, a coterie of managers, and maybe a number of executives gather in a “war room.” Up to this point, there has been minimal collaboration between development and operations. However, now these two groups are working together as a single team. They start ticking through a long checklist or playbook of manual steps.

However, even exhaustive checklists do not guarantee a problem-free deployment. Given the number of changes in the release, the deployment is likely complex and risky. As we will see in the following chapters, dependency management is challenging and “dependency hell” can be very real. So, the team might find that a key dependency was missing from the production environment. The team might discover that an incompatible library version was installed, or that a critical setting was misconfigured, or that a migration step fails or is forgotten, or that changes have caused requests to a partner service to fail.

Any number of missteps could take an already complex deployment off track. Tensions would rise, firefighting would ensue, and the hours would stretch on. The team would hope to wrap up deployment ...