AI Companies Scraped the Web; Their Next Target Is Data Showing How Humans Actually Work
A market for workplace records
AI firms are pursuing company records, including anonymized Slack messages, emails, presentations, and internal documents. Offers can reach six or seven figures, according to a report by Charles Rollet, Tom Carter, and Stephen Council.
The aim is to help AI systems learn how people collaborate, make decisions, solve problems, and carry out tasks with multiple steps. Such workflows are increasingly important to the development of AI agents.
Law firms turn to experience
A report by Melia Robinson examines law firms developing AI systems using decades of legal knowledge. Rather than depending only on ready-made AI models and tools, firms are drawing on their own expertise. That approach can make a firm's accumulated experience a competitive advantage as it builds AI for legal work.
Teaching robots physical tasks
AI companies are also paying people to record practical work, including cooking meals and packing warehouses. Rya Jetha's report describes how those recordings can teach robots how people carry out everyday physical tasks.
The different efforts reflect a broader change in AI development. After an initial period focused largely on larger models and faster chips, attention is turning toward helping machines understand how work is done.
Key points
- AI firms are seeking anonymized company records, with some offers reaching six or seven figures.
- Workplace data could teach AI agents to manage collaboration, decisions, and multi-step tasks.
- Law firms are building AI systems around their own legal experience.
- Recordings of everyday work are being used to train robots on physical tasks.