Google: $10 million. Mercor: $7.5 million.
Those bids appeared in an August 14 bankruptcy auction notice for Spirit Airlines’ deidentified corporate data. Google was selected as the successful bidder; Mercor was the backup. The notice made the transaction subject to court approval. It is evidence of named buyers bidding for a defined asset, not evidence that a sale had closed. The filing is here.
For anyone following the economics of AI, that is a revealing document. A company’s accumulated working life has become something another company is willing to price.
On October 9, micro1 announced a commitment to spend $1 billion over the following twelve months acquiring enterprise operational data. Its stated purpose is to build reinforcement learning environments that reflect real business activity. That is a prospective procurement commitment, not $1 billion already spent or a measurement of this market’s revenue.
Still, the direction is clear enough to investigate: internal workflows are attracting buyers.
What is enterprise workflow data?
Enterprise workflow data is the connected record of how an organization gets work done: messages, documents, tickets, revisions, decisions, and outcomes. Licensing it for AI means granting defined rights to use an agreed dataset, potentially after substantial filtering and transformation.
Consider the difference between a finished project report and the evidence of how that report came together. The latter may reveal conflicting instructions, failed attempts, a decision to escalate, and the correction that finally worked. For an AI agent expected to operate inside a business, those relationships can be useful training or evaluation material.
My investment question is where the durable value will accumulate: access to the records, the work of making them usable, or the environments that let a buyer repeatedly test and improve an agent.
Below: a map of four companies, a worked example of turning an archive into a testable task, and five questions I would ask before underwriting this business.
Semi Diffusion examines the business behind AI and expert work. Paid membership includes the full archive—including The Hidden Business of Making AI Fail—and access to clawded.ai for personalized AI work recommendations.