Have you heard of the Agentic SDLC? I first stumbled across the term in a Medium article recently, and it really stuck with me. It got me thinking about how we can actually prepare if this becomes the norm in the next few years.
Usually, we think of AI as a tool for quick snippets or boilerplate. But in an Agentic SDLC, AI agents aren’t just assistants. They become active participants. Imagine specialized agents working together: one researching API requirements, one drafting logic, and another checking for security gaps. You’re still a builder, but you’re also the conductor supervising how everything fits together.
The big challenge for us in low-code is the Metadata Wall. Since our logic lives inside visual models, it’s not as readable or editable for AI as plain text code.
I’ve put together some thoughts on how we might bridge this gap even if the metadata becomes more open: a staging area where agents can experiment and pre-test logic without touching the core app.
I wanted to share this here first and get your feedback:
Do you see this kind of sandbox becoming necessary?
If the metadata wall comes down, would you trust AI changes without a logic sandox first?
Full write-up here: https://medium.com/@michael.de.guzman/the-metadata-wall-why-the-agentic-sdlc-is-different-in-low-code-83e51fc63505
hello @Michael de Guzman
This is a thoughtful and timely perspective. The Agentic SDLC shifts AI from “productivity aid” to an active system actor, which fundamentally changes governance expectations—especially in low-code environments. I strongly agree that a sandbox or staging layer becomes essential, not optional. When agents can reason, propose, and modify logic autonomously, isolation is the only safe way to validate intent, performance, and side effects before promotion.
Even if the metadata wall becomes more open or machine-readable, blind trust in agent-generated changes would be risky without deterministic validation gates. A supervised sandbox aligns well with CI/CD principles and preserves human accountability while still unlocking agentic speed and scale.