The multiplier with AI: automated end-to-end tests
Nobody in this roundup mentioned automated testing, especially automated end-to-end (E2E) testing. That's not a criticism. It just reflects how the question was framed, and how testing tends to be treated in practice: as a nice-to-have that gets deprioritized under delivery pressure.
For a long time, it has been difficult to show the value of tests and prove their value exceeds their cost. I'd argue it's as important as the other practices in this article and it amplifies their impact.
Every practice described above is oriented toward shipping confidently:
Distributed local environments
CI for schema sync
Early QA for stakeholder validation
Disciplined update cadence
Many testimonials focus on the importance of automation. Automated E2E tests are what make project update confidence quantifiable. Without them, "confident" means "we reviewed the PR and it looked right." With them, it means "we reviewed the PR, it looked right, and our test suite passed."
This distinction becomes critical as soon as AI agents enter your development workflow.
AI writes code quickly. It refactors across files, resolves breaking changes, updates dependencies. But it doesn't know what it broke if it doesn't have a tool to validate it.
I wrote about exactly this scenario in Virtual Inbox, Real Tests: AI-driven E2E automation for Xperience by Kentico membership flows: an AI agent updating a membership flow that didn't touch the email template directly, but broke the variable reference that the template depended on. No compiler catches that. No unit test covers it. Only an E2E test that actually sends and receives email would have caught it before it reached production.
Thankfully, AI agents aren't just capable of quickly writing application code. They are also proficient at writing tests. If our projects are designed with automated E2E testing in mind, the cost of adding them today has been significantly reduced.
Xperience by Kentico's CI feature is also part of what makes E2E tests less brittle than they typically are in other platforms.
Developers have full control over code, content, and configuration in their local environment, so tests do not break because someone edited a page in a shared database. The environment is deterministic. This again lowers the cost of automated E2E testing.