The bankruptcy fire sale just revealed what your company's Slack channels are actually worth.
The Summary
- Spirit Airlines sold its internal corporate data — wikis, emails, source code, spreadsheets — to Google for $10 million, with AI training company Micro1 countering at $12.5 million after the auction closed
- This isn't about customer lists or flight data. It's about how work actually happens: the messy, undocumented reality of running an enterprise
- Companies building AI agents are paying top dollar for real-world business datasets because synthetic data can't replicate the chaos of how humans actually collaborate
The Signal
Spirit Airlines hasn't had a plane in the air since May. But Google just paid $10 million for something more valuable than the fleet: the digital exhaust of how 6,000 employees ran a discount airline. Not the polished outputs. The internal wikis. The email threads where someone finally explained how the scheduling system actually works. The spreadsheets held together with formulas written by an intern in 2019.
AI companies are in a land grab for enterprise datasets because training on synthetic data or public web scrapes doesn't teach models how work gets done. It teaches them how work gets documented, which are two different things. The real knowledge is in the Slack channel where ops and finance hashed out a workaround. The commit messages. The quarterly planning docs that never made it to the intranet.
"Companies are sitting on decades of records that show how real work gets done, and that data is now some of the most valuable material for training and evaluating AI."
This explains why Micro1 came back with a $12.5 million counteroffer after the auction closed. At that price, Spirit's bankruptcy estate is selling its institutional memory for roughly $2,000 per employee. That's cheap for what amounts to a training dataset on enterprise operations in a complex, regulated industry.
What makes this data valuable:
- Context-rich communications showing how decisions actually flow through an organization
- Real-world examples of how humans route around broken systems
- Edge cases and exceptions that synthetic data can't generate
Google says any personally identifiable information will be processed by a third-party "deidentification agent." The dataset won't include customer lists. But Spirit's flight attendant union has objected, arguing employees can still be identified even after names are stripped. They're right. An email thread about a specific flight delay on a specific date with specific operational details doesn't need a name attached to identify who wrote it.
The privacy angle here is real, but it's also a preview of a bigger problem. Every company that goes bankrupt now has a new asset class to liquidate: the corpus of how its people worked together. That data is worth millions to AI companies building agent systems that need to understand real organizational behavior, not the sanitized version in the employee handbook.
The Implication
If your company uses Slack, Notion, GitHub, or any system that captures how work actually happens, you're creating a dataset that has market value independent of your product or service. That value crystallizes the moment you hit financial trouble. Bankruptcy courts will sell it. Acquirers will want it. AI companies will bid on it.
The flight attendants objecting to the Spirit sale are asking the right question: who owns the digital trace of your work? Not the outputs you're paid to create. The process. The collaboration. The informal knowledge transfer that happens in channels and docs. Most employment contracts don't address this because most companies didn't think internal communications were an asset they'd ever monetize. Now they are.
Watch for two things: first, data brokers setting up shop specifically to buy enterprise datasets from distressed companies. Second, employment contracts starting to address data rights around internal communications. The Spirit sale just set a price floor. $10-12 million for a mid-sized company's institutional memory is the new baseline.