Logic Apps Automation gets managed AI models, so teams skip the model setup
Microsoft lists managed AI models in Logic Apps Automation as a public preview, but the official page is nearly empty and the details sit in earlier launch material.

Microsoft has listed Managed AI Models in Logic Apps Automation as a public preview in its Azure Updates feed, dated November 17, 2026. The headline promise is that the platform supplies the models, so teams building automations do not have to provision them.
A thin notice, with the detail elsewhere
The Azure Updates entry could not be read in full when this article was written. The page returned an empty listing, so the notice itself gives no model names, prices or regions that we could verify. Everything below about the product comes from Microsoft’s earlier material on Logic Apps Automation, which is a separate and older announcement, and it should be read as background rather than as the contents of this notice.

According to Microsoft’s Tech Community post, Logic Apps Automation was introduced on June 2, 2026, at Build 2026 and runs as a public preview at auto.azure.com. In that post Microsoft says model endpoints are available out of the box inside the managed environment. The same post also lists “Hosted Models”, described as managed endpoints needing no keys or infrastructure, as coming soon. It is therefore unclear from public material which of those two states the new preview notice describes.
What teams get and what they still configure
InfoQ’s June coverage says the platform provisions and manages the vector store and AI models behind its Knowledge as a Service feature. Teams upload documents and attach a knowledge base to an agent, and the platform handles ingestion, chunking, embedding and retrieval. InfoQ did not name the models used.
Microsoft’s post says the platform can also work with frontier, open-source, fine-tuned or local models, and that users can connect existing Foundry agents. Managed items include compute, execution environments, knowledge services and isolated sandboxes for agent harnesses. Items listed as coming soon include custom container images for sandboxes, inline Python, VNet support with private endpoints, and custom connectors. InfoQ’s account of network isolation is more positive than Microsoft’s own list, so teams with strict network requirements should check the current documentation.
Who can try it, and at what price
Per Microsoft’s June post, the preview launched in East Asia, Sweden Central, Australia East, North Central US, UK South, Southeast Asia and West US, with more regions to follow. Pricing is consumption based, with no annual commitment or per-seat license. Charges cover a small managed-environment fee, workflow execution, and optional services such as AI model usage, knowledge, sandboxes and connector calls. Idle workflows incur no compute charges, and full pricing was described as to follow. No model pricing for this notice has been published that we could find, and a preview is meant for non-production use and testing.
A strategy of selling the whole stack as one meter
The change matters mainly as a packaging choice. In the Standard tier of Logic Apps, InfoQ reports, the same knowledge capability runs on customer-owned Cosmos DB and AI model resources. The Automation tier moves that work to Microsoft, which turns model usage into another line on a single consumption bill instead of a resource the customer owns and secures.
That fits a pattern in Microsoft’s other November announcements. At Ignite the company pushed agents and Copilot deeper into its own stack, as covered in our report on Copilot Workspaces, an Autopilot agent and Cobalt 200. Managed models in a low-code automation product extend the same idea to business teams without an integration developer, which is how InfoQ describes the target audience.
The tradeoff is control. Customers who wanted to choose and govern their own model resources have that option in Standard, while Automation removes the setup work in return for less visibility. The notice does not say which models are used, how they are updated, or how usage is billed, and Microsoft’s claims about convenience have not been independently verified.
Sign Up for the Daily AI Pulse
One email a day. All the stories that matter.
By signing up, you agree to our Terms of Use and acknowledge the data practices in our Privacy Policy. You may unsubscribe at any time.
Leo MartinsAI Tools, Prompts & Practical AI ExpertLeo Martins is the AI Tools & Practical AI Expert at Aiholics, focused on helping readers use artificial intelligence to improve productivity, creativity and everyday work.
You may also like

Azure Content Understanding adds GPT-6.1 Sol as its default document model

Microsoft folds its customer AI perks into Frontier Accelerate, with no dollar figures

Azure grew 48% as Microsoft’s $677 billion backlog funds its AI spending
