HCL is excellent and very popular, and Terraform Infrastructure-as-Code (IaC) is a great choice for defining and managing the building blocks of your cloud infrastructure—arguably the gold standard when it comes to cross-platform IaC. But so is Bicep, which is in many ways inspired by HCL. In fact, if you've been working with Terraform, we highly recommend you give Bicep a try—we are confident you will love it.

For Azure, our preference is native IaC, i.e. ARM templates and Bicep files. The key reason for choosing ARM and Bicep over the other IaC variants is that ARM templates are idempotent, so they enable deployments to produce repeatable results. And since Bicep files are converted to ARM templates before being deployed, they too do benefit from this major differentiator from general, cross-platform IaC like HCL.

Our AI assistant provides as part of the solution/answer both an ARM template and a corresponding Bicep language payload to ensure any investments you've made in Bicep can still be leveraged.

One key friction point impacting the adoption of ARM templates has been the JSON-based syntax that many consider archaic (even though it is similar to AWS CloudFormation's syntax), especially when compared to HCL. Bicep solved many of those problems by providing a cleaner, more intuitive syntax. We are huge fans of Bicep.

Rather than re-invent another wheel, we address the root cause of the problem. At StratusOn, we started using generative AI internally to generate our own ARM templates. We used conversational AI and natural language to build the Quickstart Templates we offer in our Maestro Studio ENSEMBLE product. Our conversational AI journey has not been easy, but now that we've gotten the technology to a mature state—increasing the quality of its results and eliminating hallucination-type results—we are ready to share our work publicly. We are confident it will be as useful to others as it has been to us, and perhaps even more useful to many who are new to Azure and to Bicep or ARM templates.

Click here for an overview on Bicep. Also, check out the following page for an ARM template primer (great for those new to Azure)—it lists ample reasons highlighting why ARM templates are still superior and deserve to be supported: Why choose ARM templates?

Maestro Studio AI utilizes guardrails and validations that help ensure it will only provide valid, working ARM templates and Bicep files that are ready to deploy immediately.

We continue to work hard to fine-tune and improve the workflow for generating ARM templates and Bicep files from natural language queries. We have been using feedback we've received from private preview participants to improve the quality and accuracy of the results.

We've made it easy to provide us with quick feedback, from inside the app, about problems encountered or suggestions you think can help improve the product.

Yes, it supports all publicly available Azure resources.

No, we are not limiting the size of the ARM templates or Bicep generated nor the number of resources.

Registered subscribers can ask their questions in one of several natural languages, including but not limited to: English, French, Spanish, Italian, German, Portuguese, Chinese, Japanese, Korean, Vietnamese, Russian, Persian, Arabic, Hebrew and Hindi. We currently provide inside the app a set of quickstart questions in the following languages: English, French, Spanish, Italian, Arabic, Hebrew and Hindi.

The website UI is currently only available in English, but we are working on adding more languages in the near future.

While Maestro Studio AI leverages a conversational AI coding assistant that can answer any Azure infrastructure questions and provide a valid, working deployment template as a key part of the solution, we are limiting the conversation with the AI agent to a single question in this iteration of the software.

We are currently working on and plan to roll out an interactive chat experience with the AI coding agent in the next iteration, to be followed by configurable AI coding agents that are setup to monitor and automatically take action based on certain infrastructure criteria that you define.

We certainly don't think so, which is why we've waited so long before making this product publicly available.

We built this initially as something we needed to build our own ARM templates and Bicep files, and the general purpose AI coding assistants were not giving us working results—there was always something broken about 95% of the time. Just like any other AI product (or even non-AI software product), it continues to be tuned and improved—and it continues to get better, faster and more accurate. Over time, our internal project transformed into something beyond what we had originally envisioned, and we are very excited to share it publicly so that others can enjoy the same benefits.

But don't take our word for it—try it out for free and see for yourself.