Real AI Stories Project

Welcome to the Real AI Stories Project!

I am Tomas and this is my personal project I started to get deeper insights into enterprise AI. My goal is to explore how companies are using AI and how AI itself can be used to surface these insights (I know, very meta!). Right now, it is a proof-of-concept library of AI use cases and benchmarks. If you are looking for inspiration, I hope you will find it useful even in its current form.

AI Use Cases

Explore a library of around 1,300 real AI use cases collected from AWS and Google Cloud websites. Each use case has been annotated using a pre-trained AI model, allowing you to filter by industry, business function, use case taxonomy, or other dimensions. Visit the original source links to validate the information.

AI Benchmarks

Explore a collection of around 3,000 success metric benchmarks to help you calibrate your own aspirations. These benchmarks have been identified and extracted using an AI model and can be inaccurate. Refer to the original sources linked to validate the data.

Where is the project right now and where is it heading?

Currently, this is a proof of concept. To test the approach, I focused on use cases from AWS and Google Cloud websites due to their extensive coverage. Using GPT-4o-mini, I built an initial application to automate annotation and extract benchmarks. While it’s not perfect, the experiment has shown that with some structuring (e.g., taxonomies), it performs well and has potential to scale.

I plan to take the project further with a few initial ideas:

Analyze patterns and trends (e.g., common use cases by industry and function).
Expand data sources to include content from more AI companies, such as foundation model vendors, enterprise AI specialists, and vertical AI solution providers.
Enhance the use case taxonomy for improved pattern recognition.
Build an AI expert to leverage the database through RAG or similar approach.

What would be most helpful for you? Your feedback is welcome and will help shape the future of this project!

How was it built?

This proof-of-concept version has been built in Python with a focus on speed and simplicity. Here is an overview of the key features:

Repository Crawler: Uses Playwright to gather links to real-life AI use cases from AI tech companies' websites.
Content Scraper: Extracts relevant content from these websites using Trafilatura.
Automated Annotation: Leverages a pre-trained model (currently OpenAI’s GPT-4o-mini) to annotate use cases and identify success metrics.
Data Access: Baserow for easy sharing of the data.

This approach will continue to evolve as I expand the dataset and refine the insights.

Connect with me

If you would like to share feedback, ask questions, or discuss potential collaboration, feel free to reach out: