Learnings: Summarising PowerPoint presentations with AI and PowerAutomate

In the modern business world, PowerPoint presentations are indispensable for conveying information efficiently and presenting ideas impressively. In this post, I explore how the integration of artificial intelligence (AI) and Power Automate can facilitate the automated editing and summarisation of presentations. I share valuable insights, experiences and challenges that have arisen during this process.

Project objective: Automated SharePoint summaries

The main objective of our project was to use Power Automate to automatically process PowerPoint presentations and transfer concise summaries to a SharePoint list. However, this process proved to be more complex than initially anticipated.

Problems with text recognition
A major obstacle was the lack of a direct function in Power Automate to extract content from PowerPoint files. Our solution was to export the presentations as a PDF and use the AI Builder action „Recognize text in an image or a PDF document“. Despite the extraction, the result was chaotic as every single letter was captured. Many slides contained irrelevant information such as screenshots or sponsor logos, which made it difficult to create meaningful summaries.

Search for effective solutions

To improve the results, we tested the „Async Extractive Summarisation (2022-10-01-preview)“ action from Azure Cognitive Services. Unfortunately, these results were also unsatisfactory, as the summaries generated were not sufficient and the „Key Phrases (V3.0)“ function did not offer any added value in terms of keywords.

AI technology and PowerPoint: progress needed

Our experiment clearly shows that the AI technology currently available for the automatic creation of PowerPoint summaries is not yet fully developed. None of the tools tested were able to fulfil our requirements. Other methods or manual adjustments are needed to generate high-quality summaries.

Conclusion

Our plan to use AI and Power Automate to create PowerPoint summaries did not lead to the desired results. It is important to recognise the limitations of automation. Manual intervention or alternative strategies are often required, and human judgement is still crucial.