A cloud-based workflow automation diagram routes documents and requests through an AI model to automated outputs.
Microsoft has put Managed AI Models in Logic Apps Automation into preview. It announced the feature on November 17, 2026, in its Azure Integration Services blog post covering Azure Logic Apps news from Ignite 2026. The pitch is simple. Managed model endpoints are provided for you, with no keys or infrastructure to bring. That wording comes from the Build 2026 launch post for Logic Apps Automation. The Ignite post's own phrasing is that the capability offers ready-to-use AI features without developers provisioning or managing model infrastructure.

This is a short announcement, so this article is short too. Microsoft has said little beyond the headline claim, and I'll flag the gaps below.

A cloud-based workflow automation diagram routes documents and requests through an AI model to automated outputs. What Microsoft actually announced​

Microsoft sorts the AI tools in Logic Apps Automation into three tiers:

  • Workflows handle predictable tasks.
  • Decision actions are in public preview. They use a model called Microsoft-Decision-1 for focused judgments such as classifying requests, routing tickets and assessing urgency. They come as Yes/No, Choice and Score actions that return structured results.
  • Agents, including those built in Microsoft Foundry, handle more complex reasoning and action. Logic Apps can orchestrate them alongside enterprise systems and human approvals.

Managed AI Models sit beside these as a way to bring AI into workflows without running the model layer yourself. Microsoft does not say how they differ technically from Decision actions. Decision actions are a separate feature, and nothing in the announcement ties Microsoft-Decision-1 to the managed models.

Where it fits​

Logic Apps Automation is itself still in preview. Microsoft announced it at Build 2026 as a new SKU at auto.azure.com, packaging workflows, AI agents, knowledge services and model access into a managed SaaS experience. InfoQ described the intent this way: you sign in, and the compute, connectors, model endpoints and knowledge services are already there. The Ignite announcement fits that managed-SaaS direction.

The contrast with existing Logic Apps is real. Microsoft's Learn documentation for agentic workflows shows that, on other Logic Apps plans, an agent loop by default uses the Azure OpenAI model available in your logic app's region. Other documented routes need more setup. Foundry Agent Service, for example, requires a Foundry project connected to a deployed Azure OpenAI model. Managed AI Models appear aimed at removing that kind of provisioning step on the Automation SKU. Microsoft's announcement stops short of confirming exactly how.

What the announcement does not say​

Microsoft's post leaves several things open. Anyone evaluating this should treat each as an unanswered question:

  • Models: which models or providers are included, and whether you can choose between them.
  • Supported tasks: which actions or AI tasks the feature covers.
  • Region availability: where it can run.
  • Pricing and quotas: the billing unit, usage limits and cost. Separately, the Logic Apps Aviators newsletter for November 2026 says Microsoft has announced pricing and free grants for Logic Apps Automation, currently in public preview. That covers the SKU, and I did not verify whether it covers managed-model usage.
  • Data handling, residency, security controls and SLAs: none are described.
  • Setup steps: there are no portal steps, code samples or API details.

"Without managing model infrastructure" is a narrow claim. It doesn't mean no governance, no cost, or no need for AI expertise. Reading it that way would go beyond what Microsoft said.

Preview status​

Microsoft's Azure Updates site defines "In preview" as available to Azure customers for non-production use and testing. That quote is from the Azure Updates status definitions, and I could not confirm the specific update entry on that page. Treat this as an evaluation feature. It is not something to put in front of production business processes yet.

Practical takeaways​

  1. Try it in a sandbox. If your team already uses the Automation SKU, test managed models on low-risk workflows.
  2. Ask the compliance questions early. Data residency, logging and model behavior are unanswered, so ask them before any real data goes in.
  3. Compare against your current approach. If you now maintain your own model endpoints for Logic Apps, weigh the lower setup effort against the control you give up. That trade-off is general industry reasoning, not something Microsoft stated.
  4. Wait for documentation. Pricing, supported models and regional details should come with fuller docs before any rollout planning.

Bottom line​

Managed AI Models fits Microsoft's push to make Logic Apps Automation a turnkey environment for AI-assisted workflows. The announcement is credible, but it comes from Microsoft alone, with no independent testing or performance data. Until the missing details arrive, it is a promising preview and not yet a planning basis.

 

References

  1. (In preview) Public Preview: Managed AI Models in Logic Apps Automation Azure Updates 2026-11-17T20:00:32Z
  2. Microsoft Launches Logic Apps Automation at Build 2026 - InfoQ infoq.com
  3. 🎉 Automation just became a team sport. Meet Azure Logic Apps Automation. techcommunity.microsoft.com