The value of wood shavings.

There is a recent news story that, in my view, is worth more than many essays on the data economy. Startups that shut down are selling their work archives to AI labs: Slack, emails, documents, calendars, code. One founder reports receiving around $25,000 for the code and an estimated additional $15,000 for data from workplace tools.

The scale is growing fast. According to the article, SimpleClosure has completed more than 150 deals and expects to return over $10 million to startups by the end of 2026. Now, 47% of its liquidations include the sale of code or data. In the Spirit Airlines bankruptcy, Google bought decades of corporate data for $10 million: 100 million emails, 500 million Teams messages, 176,000 employee records.

Particleboard and tacit knowledge

The founder of Sunset, another company in the sector, uses an image I find illuminating: particleboard. These “operational waste” data are like wood shavings that everyone considered worthless but that actually have tremendous value.

Why are they valuable? Because labs have exhausted public data, and synthetic data are showing their limitations. Training agents capable of working requires real traces of work: disagreements, difficult decisions, ambiguous judgments, corrected mistakes. An executive at Surge AI explains that a company’s Slack history, if rich and varied, can be worth more than another company’s Slack, emails, and calendars combined.

This is the key point for anyone who, like me, has long been thinking about the cognitive enterprise. The value does not lie in formal documents, procedures, or organizational charts. It lies in the tacit knowledge conveyed through conversations: how decisions are really made, how exceptions are negotiated, how problems that no manual anticipates are solved. Someone has just put a market price on this knowledge.

Three open questions

• Who owns the value. The proceeds go to creditors and investors, not to the workers who produced that knowledge. Ifeoma Ajunwa, a professor at Emory University, speaks of “captured capital” and calls for employees’ explicit consent. In the United States, there is no federal law prohibiting these sales.
• How clean the data are. One lawyer describes them as “legally radioactive.” Personal data and HR files are removed, but medical appointments, children’s school pickup times, and private lives remain in emails and calendars. In Europe, under the GDPR, a transaction like the Spirit Airlines deal would raise serious questions about legal basis and purpose.
• What they are really worth. No one has yet published tests comparing training on real data with the alternatives. Prices are based on “research theses.” It is a market emerging before its metrics.