
This companion guide is designed for developers, architects, and administrators who have recently reviewed Paul Stork’s foundational 2024 session on Microsoft Power Apps Copilot. The artificial intelligence landscape within the Microsoft Power Platform has evolved at a staggering pace since that recording. As of March 2026, Microsoft has fundamentally shifted away from treating Copilot as a novelty embedded widget that you drag onto a specific screen. Instead, Copilot is now a native, ambient intelligence layer infused directly into the application framework, the Power Apps Studio, and the broader Microsoft 365 ecosystem.
You are no longer merely building applications with an AI assistant; you are building agentic applications where Microsoft 365 Copilot, Agent 365, and Microsoft Dataverse act as a unified, seamless enterprise ecosystem. The underlying models have been upgraded to the GPT-5.3 Instant and GPT-5.4 Thinking variants, offering vastly superior reasoning, context windows, and generation speeds. Furthermore, the introduction of the Microsoft 365 E7 "Frontier Suite" has completely redefined how we license, govern, and deploy these AI capabilities securely across large organizations.
This comprehensive report bridges the gap between the isolated, component-based techniques demonstrated in the original 2024 session and the modern, enterprise-grade, agentic approaches required for production workloads today.
What's Changed Since This Session
The table below provides a strictly factual summary of the core technologies demonstrated in the 2024 session, their current availability status as of March 2026, and the modern architectural replacements you must use today.
| Technology / Demo Feature from 2024 Session | Status in March 2026 | What Replaced It Today |
|---|---|---|
| App Generation via pure Text Prompt | Evolved | Start with Data Workspace. AI forces a data-first approach, generating the Dataverse table in a dedicated workspace before building the canvas UI. |
| Copilot Control (Embedded Chat Widget) | Deprecated (Feb 2, 2026) | Microsoft 365 Copilot natively in Canvas Apps. Integrated directly into the app header. Requires no screen real estate or manual layout changes. |
| Copilot Answer Control (Clickable queries) | Deprecated (Feb 2, 2026) | Prompt Guide Customizations. Pre-built queries are now injected directly into the native Copilot chat pane using Power Fx payloads and Copilot Studio. |
| Custom Copilot Component | Deprecated (Feb 2, 2026) | Agent 365 & Copilot Studio Native Integration. Custom agents run via the universal Microsoft 365 Copilot fabric rather than isolated canvas controls. |
| Formula Bar Copilot Builder | Evolved | Comment-Generated Formulas. You can now type standard code comments (// or /*) directly into the formula bar to generate Power Fx. |
| Formula Explanation | Enhanced | Explain this Selection. You are no longer restricted to explaining an entire formula; Copilot can now explain partial highlights of complex legacy code. |
| Underlying AI Model | Upgraded | GPT-5.3 Instant & GPT-5.4 Thinking. Vastly superior reasoning, context window, and generation speed compared to legacy 2024 models. |
Architectural Shift: From Embedded Controls to Native Shell Integration

As of February 2026, the drag and drop Copilot Control is deprecated. Natural language interaction is now handled natively by the Microsoft 365 Copilot pane, preserving canvas real estate and providing deeper app context.
How to Build This Today
The original session demonstrated seven core scenarios that introduced makers to the concept of generative AI in low-code environments. However, Microsoft has completely overhauled the user interface, aggressively deprecated legacy controls, and introduced the GPT-5 class of models to drive agentic workflows. Here is exactly how you execute those same seven scenarios today in March 2026, using the latest supported methodologies and official enterprise patterns.
Demo 1: Creating Canvas Apps from Scratch
The session showed Paul Stork using a single text box on the legacy Power Apps home screen to generate a fully functional application and its underlying Dataverse table in one seamless, but slightly opaque, click.
Here's how to build that same thing today in March 2026. Microsoft has intentionally decoupled the data modeling phase from the app generation phase to ensure better enterprise architecture. You now use the dedicated Start with Data workspace. This forces developers to interact with and validate the AI-generated schema before the app UI is built, preventing messy, unscalable data structures from entering your environments.
Step-by-Step 2026 Guide:
- Navigate to the modern Power Apps home screen (
make.powerapps.com). - Select the Start with data tile from the primary dashboard.
- Select the Create new data option.
- The system will load the Create new tables workspace. You will see a blank grid representing a table in the center, and the persistent Copilot panel docked on the right side of your screen.
- In the Copilot panel text box, enter your natural language prompt. For example, if you are building a hotel management tool, type: "Create tables to track hotel housekeeping tasks including room numbers, task types, staff assignments, and task status".
- Press Enter or select Submit. The AI, powered by the highly responsive GPT-5.3 Instant model, will generate the schema in seconds.
- Crucial 2026 Step: Do not instantly click save. Use the conversational panel to iteratively refine the schema. You can issue secondary commands such as, "Change the task status column to a choice field with options: Pending, In Progress, Completed."
- You can also instruct Copilot to build relational data models right in this view by asking it to relate the new table to an existing Dataverse table like Accounts or Contacts.
- Once you have thoroughly validated the schema, click the Save and exit button. Power Apps will now automatically generate the standard responsive canvas app connected to this newly minted, validated Dataverse table.
Quick Win: Do not skip the mock data generation phase. In 2026, the Copilot panel allows you to view and manipulate sample data before the app is generated. Asking Copilot to "Add 15 rows of highly realistic sample data to the tables" will make your canvas app look fully populated and ready for immediate stakeholder demonstrations the moment it finishes compiling.
Demo 2: Creating Tables from Scratch Using Copilot Inside the Editor
The session showed Paul Stork creating a new Dataverse table directly from inside the Power Apps Studio editor by asking Copilot to add a data source on the fly.
Here's how to build that same thing today in March 2026. This workflow is now much tighter and significantly more resilient. The integration of Copilot with Dataverse allows you to engineer complex, localized relational tables without ever having to break your flow by opening the backend Power Platform Admin Center.
Step-by-Step 2026 Guide:
- While actively working inside your open Canvas App in the Power Apps Studio, navigate to the left-hand authoring menu.
- Select the Data icon (represented by the cylinder symbol).
- Click + Add Data and then select Create new table.
- The familiar Copilot table generation interface will appear as an overlay modal directly on top of your canvas workspace.
- Type your prompt to define the table structure.
- Leverage the new 2026 refinement commands to enforce data integrity. You can instruct Copilot to configure advanced column behaviors directly from the prompt. For example, type: "Make the Email column mandatory, ensure it only accepts valid email formats, and set the default value for the Status column to Active."
- Review the proposed table structure in the preview grid.
- Click Create. The table is instantly provisioned in your environment's Dataverse instance, and the data connection is automatically injected into your current canvas app's data sources.
This localized table creation is particularly useful when you are building modular canvas apps that require highly specific, app-scoped data tracking alongside your main enterprise data sources. It allows citizen developers to safely extend data models without requiring extensive backend database administration skills.
Demo 3: Using Copilot to Modify an Existing App
The session showed Paul Stork using a floating chat window to ask Copilot to add screens, insert galleries, and format controls on an existing canvas app.
Here's how to build that same thing today in March 2026. The Edit with Copilot side pane is now the central command center for all maker activities. Microsoft has standardized this right-aligned pane across all Microsoft 365 and Power Platform applications to ensure a consistent developer experience, improving workflow and accessibility. Furthermore, Copilot is now aware of modern controls and enterprise theming.
Step-by-Step 2026 Guide:
- Open your existing Canvas App in Power Apps Studio.
- If the Copilot pane is not open by default, navigate to the top command bar and click the Copilot button to reveal the right-hand Edit with Copilot pane.
- Select any control or screen on your canvas. The Copilot pane gains immediate context regarding exactly what you have selected, leveraging the new Power Apps Model Context Protocol (MCP).
- Type your command to manipulate the UI. For example, select a gallery and type: "Change the layout of this gallery to show the profile image on the left and the primary text on the right."
- New for 2026 - Theme Application: You can now use Copilot to manage global app branding via natural language. You can type, "Apply the corporate blue theme to all primary buttons on this screen." Copilot can now interpret and apply YAML-based styling tokens across your app portfolio, ensuring visual consistency.
- New for 2026 - Modern Control Swaps: Copilot natively understands the nine new modern controls (Combo Box, Date Picker, Tab List, Info Button, etc.) that received massive quality updates in the March 2026 wave. You can ask Copilot to "Swap this legacy dropdown for a modern Combo Box" and it will execute the replacement while maintaining the underlying data mappings.
Quick Win: Look out for the new Draft with Copilot feature designed for your end-users. When you add a modern rich text editor or multi-line text input to your canvas app, users will automatically see a Copilot icon allowing them to draft, summarize, or rewrite text inside the app. This requires zero extra configuration or coding from you as the maker.
Demo 4: Generating Power Fx Formulas from Comments
The session showed Paul Stork using Copilot in the function bar by clicking an icon, opening a specific builder window, and typing a descriptive comment to generate a Power Fx formula.
Here's how to build that same thing today in March 2026. Microsoft has aggressively streamlined this interaction. You no longer need to open a separate, clunky builder window. The intelligence is baked directly into the primary formula bar, recognizing standard developer syntax natively. This is essentially the GitHub Copilot experience, brought natively into the low-code ecosystem.
Step-by-Step 2026 Guide:
- Ensure the feature is active for your environment. Go to Settings > Updates > Preview tab within your app and verify that the Copilot for formulas or Copilot comment-generated formulas toggle is set to On. Note that this specific feature is built-in and does not require an additional Microsoft 365 Copilot license.
- Select a control on your canvas (e.g., a Text Label) and select the property you want to modify from the property dropdown (e.g., the
Textproperty). - Click directly into the main formula bar at the top of the screen.
- Type a standard code comment using either the
//or/*syntax. - Immediately type your natural language request. Example:
// Calculate the variance between the target revenue and actual revenue from the Financials table and format the result as currency. - Press Enter (or simply wait a few seconds). Copilot will parse your comment, query the context of your data connections, and generate the exact Power Fx code directly beneath your comment, formatted as a ghosted gray text suggestion.
- Press the Tab key on your keyboard to accept and implement the suggested formula.
The beauty of this 2026 approach is dual-purpose. Not only is the generation faster due to GPT-5.3 Instant, but your original prompt (// Calculate the variance...) remains in the formula bar as permanent, valuable code documentation for the next developer who opens the app.
If you prefer a more discrete UI interaction, you can still type = anywhere in the grid or click the Copilot icon next to the fx field and select Create a formula to utilize a dedicated text-box interface.
Demo 5: Generating Descriptions for Existing Power Fx Formulas
The session showed Paul Stork using Copilot to analyze an existing block of Power Fx and generate a plain-language description of what the code does, allowing him to add it as a documentation comment to help future maintainers.
Here's how to build that same thing today in March 2026. This feature is significantly more granular and powerful. In the past, Copilot struggled with massive, monolithic formulas, often providing overly generic summaries. The 2026 update introduced the ability to isolate and explain partial code selections, which is invaluable when debugging complex legacy applications or untangling nested logic.
Step-by-Step 2026 Guide:
- Select a control with a complex, existing formula in the formula bar.
- If you want the entire formula explained at a high level, simply click the dropdown arrow near the
fxfield label and select Explain this formula. - The 2026 Upgrade: If you only want to understand a nested
LookUp, a specificPatchcommand, or a mathematical operation buried within a massive block of code, use your mouse to highlight just that specific segment of text in the formula bar. - Once highlighted, click the dropdown near the
fxlabel and select the new Explain this selection option. - Copilot utilizes its deep context window—powered by the advanced reasoning capabilities of the GPT-5.4 Thinking model—to understand not just the isolated highlighted code, but the surrounding context and data dependencies. It will return a highly specific, plain-English explanation of that exact function.
- You can then copy this explanation and paste it directly above the code block with a
//prefix to permanently document the logic for your team.
Quick Win: Always use the Explain this selection tool when working with User Defined Functions (UDFs). UDFs are a massive architectural addition to Power Apps in 2026, allowing you to write a complex formula once and reuse it across the entire app. Documenting the core logic, inputs, and outputs of your UDFs with Copilot ensures your application remains highly maintainable as it scales.
2026 Copilot Implementation Decision Matrix

Follow this logic to implement user-facing AI in your 2026 Power Apps. Legacy embedded controls are retired in favor of native header integration and enterprise-governed agents.
Data sources: Microsoft Learn (Copilot Answer Control), Microsoft Learn (Modern Controls), Microsoft Learn (Add AI Copilot), Microsoft Learn (Release Plan), Microsoft Learn (Prompt Guide)
Demo 6: Embedding a Chat Box for Data Queries (The Old Copilot Control)
The session showed Paul Stork dragging a "Copilot Control" component from the insert menu directly onto the canvas app screen. This created a dedicated, hard-coded chat box widget where users could ask questions specifically about the app's Dataverse records.
Stop right here. This is the single most important architectural update you must understand in this 2026 guide.
As of February 2, 2026, Microsoft officially deprecated and discontinued the addition of the embedded Copilot Control to all new canvas apps. Existing apps containing the control will remain functional for a very limited grace period, but they are completely unsupported moving forward.
Dragging a chat widget onto a screen was an inefficient use of mobile and desktop UI real estate, and it siloed the AI's context. Today, you must transition to Microsoft 365 Copilot in canvas apps. This provides a native, unified, and globally accessible experience. The AI now lives in the overarching application shell (the header), allowing users to open a chat pane alongside their work without navigating to a specific screen that happens to contain a widget.
Step-by-Step 2026 Guide to Enable Native Chat:
- A Power Platform administrator must first ensure the foundational tenant-level setting is active. In the Power Platform admin center, navigate to your specific environment. Go to Settings > Product > Features and ensure the toggle for Allow users to analyze data using an AI-powered chat experience in canvas and model-driven apps is set to On.
- Open your canvas app for editing in the Power Apps Studio.
- Navigate to the top command bar and select Settings > Updates.
- Click on the Preview tab.
- Locate the setting for App Copilot (often labeled as Microsoft 365 Copilot chat depending on your deployment ring) and toggle it to On.
- Save and Publish the application.
- When your end-users launch the app in a web browser or the Power Apps mobile client, they will not see an embedded widget blocking their data. Instead, they will see a native Copilot icon persistently available in the global app header.
- Clicking this header icon slides out a persistent right-hand pane. Here, users can interrogate app data, generate visualizations using the code interpreter, and even take actions across the broader Microsoft 365 ecosystem—such as drafting documents or scheduling meetings based on the app's context—without ever leaving the workflow.
This shift to agentic workflows means Copilot is no longer restricted to just passively reading Dataverse rows; it can actively take action, generate cross-platform documents, and trigger complex automations seamlessly.
Demo 7: Adding Clickable Standard Queries (The Old Copilot Answer Control)
The session showed Paul Stork using the "Copilot Answer Control," which was a brand-new feature at the time. This was a specific UI component, heavily optimized for mobile users, that provided end-users with pre-built, clickable buttons containing standard queries (e.g., "What are my pending tasks?"). It allowed for one-click AI responses without requiring the user to type a prompt.
Stop right here. Just like the primary Copilot Control discussed in Demo 6, the Copilot Answer Control was completely discontinued on February 2, 2026.
Microsoft realized that hard-coding predefined AI questions into the physical app layout defeated the purpose of a flexible, natural language interface, and required constant app updates whenever a new common question arose. Today, you achieve this exact same user convenience by customizing the Prompt Guide directly within the native Microsoft 365 Copilot pane. You do this using Power Fx payloads and Microsoft Copilot Studio, moving the configuration from the UI layer to the agentic logic layer.
Step-by-Step 2026 Guide to Build Pre-built Queries:
- Ensure your canvas app is connected to a custom agent or Copilot Studio backend.
- Open the backing agent in Microsoft Copilot Studio and add a new blank topic.
- Rename the topic to accurately reflect its intent (e.g., "App Quick Prompts" or "Dashboard Navigation").
- Change the topic trigger to Event received.
- Under the Event received configuration, set the exact event name to
Microsoft. - Under this event trigger, click the + icon to add a new node, and select Variable management > Parse value.
- In the Parse value box, you will inject a specific JSON payload using Power Fx. This payload defines the clickable buttons (now called "sparks") that will appear contextually when the user opens the native Copilot pane.
- Paste the following formula architecture, modifying the
displayNamefields to match your desired queries:
Code snippet
}]
- Select Insert, save your topic, and publish your agent.
Now, when your users click the ambient Copilot icon in the global app header, they will be greeted with your custom, clickable prompts ("sparks") sitting natively and cleanly in the chat interface. You achieve the exact same user experience Paul Stork demonstrated in 2024, but with zero layout disruption, dynamic update capabilities, and full alignment with 2026 enterprise architecture.
Alternative Quick Win: If you absolutely must have a physical button directly on the canvas screen that triggers an AI response (for example, a "Summarize This Record" button next to a form), you should use AI Builder Custom Prompts. You can build a prompt in the AI Hub, attach it to your app data, and link it to a standard Power Apps button control. When the user clicks the button, it runs the prompt in the background and outputs the generative result to a standard text label.This avoids using deprecated controls while still delivering one-click AI.
Licensing Quick Reference
As Copilot matures from a localized maker novelty into a foundational, enterprise-wide workflow tool, the licensing models have grown substantially more structured. In 2026, administrators and architects must carefully navigate between App execution licensing, AI model compute credits, and the new comprehensive agent governance suites.
The table below outlines the core licensing requirements and critical context for the features discussed throughout this guide.
| Feature / Capability | Required License(s) in March 2026 | Important Context for 2026 |
|---|---|---|
| Copilot Chat in Standalone Power Apps | Power Apps Premium ($20/user/mo) OR Power Apps per app ($5/user/app/mo) OR Pay-as-you-go. | Gives end-users access to run the custom canvas app and utilize the native M365 Copilot chat capabilities within it. |
| Copilot Chat in Model-Driven Dynamics Apps | Dynamics 365 (Sales Enterprise, Premium, Customer Insights, Field Service, Project Operations). | Copilot chat and form fill assistance are generally available and turned on-by-default for properly licensed D365 users. |
| Building Custom Agents (Copilot Studio) | Copilot Studio User License ($0/user/mo for authoring). | The authoring license for makers is free, but deploying and running these agents in production consumes tenant-level Copilot credits. |
| AI Builder & Custom Prompts | Consumes Copilot Credits or AI Builder Credits. | Critical Warning: As of late 2026 (November 1), legacy AI Builder credits included in per-user plans are being completely phased out in favor of unified Copilot credits. Plan your capacity accordingly. |
| Enterprise Scale & Agent Governance | Microsoft 365 E7 (Releasing May 1, 2026 at $99/user/mo). | The new "Frontier Suite." It bundles M365 E5, Entra Suite, M365 Copilot, and Agent 365. Agent 365 is absolutely crucial for governing custom Copilots across your entire Power Platform securely. |
| Advanced Security (CMK & Lockbox) | Managed Environments + A5/E5/G5 Level Subscriptions. | Customer Managed Keys (CMK) and Customer Lockbox enforcement require activation of Managed Environments and top-tier M365 licensing. Copilot natively respects these boundaries. |
Implementing Microsoft Power Apps Copilot in 2026 requires strict adherence to these new architectural and licensing paradigms. By aggressively moving away from legacy embedded canvas controls, embracing the native M365 Copilot shell, leveraging the GPT-5 model family, and deploying Agent 365 for governance, your applications will be faster, vastly more resilient, and fully aligned with Microsoft's long-term product roadmap.