0–N scaling on dedicated compute, the same 1,400+ connectors, and a new designer at auto.azure.com. That’s the core of Azure Logic Apps Automation, now in preview. Why care: if you’re building agentic, variable workflows, this points to an AI‑first path without standing up always‑on capacity.
What it is
A new SKU on the Logic Apps runtime you already know (same connectors, same engine), with an AI‑centric authoring model and a new management hierarchy (projects → applications → workflows). There’s a separate portal and designer, aimed at dynamic, agent‑style orchestration. Docs position it alongside Standard and Consumption rather than a replacement.
What stood out in the hands‑on
- New canvas feels faster; vertical/horizontal layouts help. Real‑time run history is actually useful for debugging.
- Early gaps: the post finds no end‑to‑end CI/CD yet and no full project export; copying workflow JSON from the portal works but is not the same as proper source control. Model availability is inconsistent in preview; some models failed while others worked.
- Networking and pricing are still the big question marks. Docs talk about enterprise isolation; in preview, verify in your region/tenant what’s actually lit up before you commit.
Where this fits
- Use Automation to prototype agentic, decision‑heavy workloads and variable traffic you want to scale to zero.
- Keep Standard for deterministic, high‑throughput integrations and B2B/EDI; Consumption still makes sense for simple event‑driven tasks.
- The broader “agentic Logic Apps” story already exists across SKUs (see Labs). Automation mainly packages it with an AI‑first design and new operational surface.
Practical takeaway: try it for new AI/agent flows, but don’t plan a platform migration until CI/CD, network controls, and pricing are clear. Read the original for screenshots, side‑by‑side designer views, and gotchas from real testing.