AI: Real Work Data Becomes the Priority

After massively scraping the web, AI companies are now looking to collect data from real-world work. They are paying for the recording of physical tasks, purchasing internal archives for high amounts, and law firms are building their own systems based on decades of experience.

Paid Recordings of Physical Tasks to Train Robots

Some AI companies are paying individuals to record concrete activities, such as meal preparation or warehouse packing. This data is used to train robots to replicate everyday actions. This approach is part of an evolution in AI development, aiming to enable machines to understand how work is done in order to better assist humans.

Internal Archives Exchanged for Six and Seven-Figure Sums

A market has developed around corporate records: AI companies are offering sums of up to six or seven figures to obtain anonymized Slack messages, emails, presentations, and other internal documents. The goal is to expose AI systems to concrete examples of problem-solving, collaboration, decision-making, and executing complex tasks. These real workflows are used to fuel the next generation of AI agents.

Law Firms Rely on Their Own Corpora Against Off-the-Shelf Models

In the legal sector, firms are building AI systems from decades of internal knowledge, transforming their experience into a competitive advantage rather than relying solely on standard models and tools. This trend emerges as the first wave of generative AI has focused on increasing model size and chip speed, and many companies have already scraped a significant portion of publicly available web data.