The latest update to Microsoft Copilot has introduced tools that were previously reserved for software developers and data scientists. As demonstrated by creator Robert Triebwasser (@robert.triebwasser), these new “superpowers”—driven by the GPT-5.4 “Think Deeper” model—allow you to transform raw data into professional visuals with simple conversational commands.
If you aren’t “tech-savvy,” don’t let terms like “HTML” or “Data Models” intimidate you. Think of this as having a very fast personal assistant who can draft your entire report before you even touch a keyboard.
1. Switching to “Think Deeper” Mode
Standard AI is built for speed, which can sometimes lead to mistakes in complex logic. The Think Deeper mode tells the AI to “pause and plan” its answer before typing, ensuring much higher accuracy for data tasks.
-
Where to find it: Open Copilot (usually in Microsoft Teams or at copilot.microsoft.com). Look for the model selection dropdown in the top-right corner.
-
What to select: Choose GPT-5.4 Think Deeper. You will know it’s working because you’ll see a status message like “Reasoning completed in 15 steps” when it answers.
-
When to use it: Use this when you have a complex problem, like “How should I organize my budget for next year?” rather than simple facts like “What is the weather?”
2. Getting Data to Work With
You don’t need to have a perfect spreadsheet ready to try this. You can ask Copilot to create “Synthetic Data”—essentially high-quality “practice” files that look and act like real business records.
-
The Goal: To get a file you can use for practice or to show a boss what is possible.
-
What to type: “I want to practice data analysis. Please create a folder with fake data for a small business, including a list of customers, their orders, and a list of products. Send it to me as a ZIP file.”
-
What happens next: Copilot will create these files from scratch, establishing links between them (like matching a Customer Name to an Order Number).
3. Creating Your Own Interactive Dashboard
A “dashboard” is just a page with charts and buttons that update when you click them. Normally, you’d need to know how to code to make one. Now, Copilot handles the “heavy lifting.”
-
The Goal: To see your data as a professional webpage instead of just a list of numbers.
-
What to type: “Use the data we just talked about to build an interactive dashboard. Make it a single HTML file I can open in my browser. I want to see a chart for monthly sales and a ‘Top 5 Products’ list.”
-
How to open it: Copilot will give you a file (usually ending in
.html).-
Download the file to your computer.
-
Right-click it and select Open with… and choose Chrome or Edge.
-
It will open a beautiful webpage where you can click and filter your data.
-
4. Making an Editable PowerPoint
The most frustrating part of AI presentations used to be that you couldn’t change the charts once they were made. This update fixes that.
-
The Goal: To get a 10-slide presentation that stays fully editable in PowerPoint.
-
What to type: “Now, take that same data and create a PowerPoint presentation for a meeting. Make sure the charts are editable so I can change the colors and numbers later if I need to.”
-
The Advantage: Unlike older AI tools that just gave you a static picture, this generates a real
.pptxfile. You can click on the bars in the charts to change the data or swap out the text yourself.
Summary Checklist
-
The “Brain”: Make sure you are in Think Deeper mode.
-
The “Body”: Ask for Synthetic Data if you don’t have your own.
-
The “View”: Ask for an HTML Dashboard to see the big picture.
-
The “Report”: Ask for an Editable PowerPoint to share with others.
By following these steps, you are letting the AI do the manual labor of organizing and designing, leaving you to simply review the results and make final decisions.
This video provides a practical walkthrough of using Copilot directly inside Excel to generate a sales dashboard similar to the workflow described above.
The Test
Click To Check It Out.
Pretty Amazing. I created some sample data using Copilot/5.4 and did some back and fort with that (thinking forward to what my questions should be.
First try, worked perfectly. The only problem I ran into was creating a bad permission of the file which I had to change on my Mac first and then copy to the webserver. I’m pretty floored.
