New and improved: Build apps, extend agents, and transform business processes in Microsoft Copilot Studio
If your business wants a simple internal app or workflow, such as an approval tracker or intake form, without hiring a developer, Copilot Studio can now generate a working draft from a plain-language description, with the underlying code visible if your IT team wants to take it further.
What changed
Source summary
Microsoft's October Copilot Studio roundup covers several additions. Makers can describe an app in natural language, get a draft they can preview/refine, and access the underlying code, in public preview, running on the new Microsoft Copilot Managed Runtime, a Microsoft-hosted runtime with identity, governed data access, and versioning for apps built in Copilot Studio, Copilot Cowork, and Copilot Code. "Hooks" add deterministic logic at points like session start, tool calls, and errors, currently in preview for agents using the GitHub Copilot harness. Foundry IQ integration lets one knowledge base serve multiple agents with citations and private networking, generally available. Expanded evaluations add testing of AI nodes and whole automations, new metrics (Task Completion, Tool Accuracy, Safety, latency), a Custom Graders Library, generated graders (tenant-specific), and an Evaluation Viewer role. A Review panel, generally available, flags publishing blockers and warnings during authoring, including when an agent has never been evaluated. A plugin registry offers a shared catalog of 100+ plugins, with Copilot Studio support still rolling out. No plan, region, or admin restrictions are stated beyond the preview/GA labels noted above.
Why it matters to your work
Practical interpretation
If your business wants a simple internal app or workflow, such as an approval tracker or intake form, without hiring a developer, Copilot Studio can now generate a working draft from a plain-language description, with the underlying code visible if your IT team wants to take it further.
When this could help
Building a lightweight internal tool (approval tracker, intake form, simple workflow) quickly, or when an IT team needs to evaluate and govern agents built by multiple makers.
Copilot Studio can now turn a plain-language description of an app into a working draft with visible underlying code, and adds governance, evaluation, and plugin-catalog features for teams scaling up agent use.
How to get started
Open Copilot Studio and describe the app or workflow you want in plain language.
Provider guidance ↗Preview and refine the generated draft; view the underlying code if you or your IT team want to extend it.
Provider guidance ↗Before publishing an agent, check the Review panel for flagged blockers or warnings, including whether it has ever been evaluated.
Provider guidance ↗
Try this · The SCALA method
Give AI a clearer brief
- Situation
- My team runs a recurring internal process using spreadsheets and email that needs a lightweight app.
- Consultant
- A low-code app builder that can draft a working internal tool from a plain description.
- Action
- Build a simple app where my team can submit a request, see its status, and get notified when it is approved.
- Language
- A previewable draft app with visible underlying code.
- Ask
- What data does the app need to capture, and who should be notified at each stage?
SCALA suggested prompt · Adapt the context and check the result.
Learn the SCALA method →Build it into your routine
SCALA suggested workflow
New internal tool
When a manual process reaches the point where it needs a lightweight app, describe it in Copilot Studio and preview the draft before deciding whether it needs developer involvement.
Agent governance
Use the Evaluation Viewer role and expanded evaluation metrics (Task Completion, Tool Accuracy, Safety, latency) to monitor agents over time once they are in regular use.
What to check
- Natural-language app building is in public preview; expect it to change and to need review before relying on it for critical processes. Source ↗
- Hooks (deterministic logic at session start, tool calls, errors) only work for agents using the GitHub Copilot harness, not all agent types. Source ↗
- Generated graders for evaluation are tenant-specific and visible only within your own tenant. Source ↗
Sources & provider guides
Official documentation checked 12 Oct 2026. SCALA prompts and workflows are suggestions, not claims of hands-on testing.
Source & availability
This is a summary of a published provider announcement, with practical context from SCALA’s feed.
Plan, region, workspace and rollout conditions may affect access in Australia. Check the original source for current availability.
www.microsoft.com