Smarter answers with embedded image understanding
[Windows]
Declarative Agents can now interpret and ground responses using embedded images in Word, PowerPoint, and PDF files. This enhancement enables Copilot to extract insights from charts, diagrams, and screenshots, improving the accuracy and completeness of answers.
Roadmap ID:
531758
Details:
What changed:
Declarative Agents now analyze embedded images in supported file types (.docx, .pptx, .pdf). Visual elements such as charts and diagrams are used to inform responses, resulting in more contextually accurate answers.
Why: Users often include critical information in visuals. This change ensures Copilot can interpret both text and graphics for more complete responses.
Try this:
Upload a Word document with embedded charts and ask Copilot to summarize key insights.
Attach a PowerPoint file with diagrams and request an explanation of the visual flow.
Use a PDF with screenshots and ask Copilot to describe the process shown.
Why this matters:
Business impact:
Improves Copilot’s ability to deliver comprehensive answers grounded in both text and visuals.
Personal impact:
Helps users get more accurate and complete responses from files with embedded images.