tech
Executive Briefing: How Microsoft, Bayer, and Discovery Use AI on the Data You Can't Upload
You have a file you would never paste into a chatbot. We all do: a contract, a board deck, a client record. AI could obviously help, but you also know you can’t send it to a model provider, so the work stays manual, or it doesn’t happen at all.
TL;DR
- Companies like Bayer and Discovery Bank are fine-tuning AI models on proprietary data to avoid sending sensitive files to external providers.
- Bayer uses a fine-tuned Microsoft Phi model for crop-protection queries, reducing resolution time from days/weeks to seconds.
- Discovery Bank fine-tuned Azure OpenAI models for tasks like understanding financial language and generating SQL, decreasing response times.
- Fine-tuned models remain exclusive to the customer and are not used to improve general foundation models without permission.
- LM Studio allows users to run AI models locally on their laptops, disconnecting from the internet to process sensitive documents securely.
- Local AI testing helps determine which jobs fit on a laptop and when an enterprise system is required.
- The article will cover offline setup, a model leak incident, the limits of laptop AI, and Microsoft's approach to model dependency.
- The goal is to enable secure AI processing of sensitive files, starting with individual documents and scaling to enterprise solutions.