Field note · Data

Your conversations are an asset.

Failed businesses are selling their old email, chat, and tickets to AI labs. The buyers aren't after answers. They want a record of how people actually get work done.

By Ryan · BeliefEngines · October 2026

A failing company used to sell its equipment, code, and customer list. Now its emails, chat messages, and work tickets get a second life.

What a shutdown sells
What a shutdown used to sell (code, equipment, customer list, brand) next to what it sells now (email, chat, work tickets, documents), all going to AI labs, data brokers, and AI startups. BEFORE NOW, ALSO Software code Equipment Customer list Brand name Email Team chat Work tickets Documents BUYERS AI labs Data brokers AI startups The buyer used to be private equity. Now it's whoever is building AI.

What a company's messages are good for

Your emails, chats, and work tickets are a written record of how your business really runs. AI can use that record two ways.

1. Rebuild the company. The messages show who does what, who they ask, and how problems get fixed. That's enough to build a practice copy of the company, where AI learns the jobs before doing them for real.2 Carnegie Mellon tested AI inside a practice company like this. The best AI finished 30% of the tasks on its own.5

What the internet teaches vs. what an inbox teaches
The internet shows a finished answer. A company inbox shows how the work got done: a customer complains, support hands it to finance, finance finds and fixes the problem, support tells the customer. THE INTERNET The answer. Not how anyone got there. A COMPANY INBOX CUSTOMER "My refund never arrived." SUPPORT → FINANCE "Can you check this one?" FINANCE "The bank bounced it. Re-sent." SUPPORT → CUSTOMER "Fixed. It lands Friday." An AI can copy the answer on the left. It learns the job from the right.

2. Fix the company you have. The same messages show where work gets stuck and who goes around whom. Microsoft studied the email, chat, and calendars of 61,000 employees and saw remote work wall teams off from each other.4 Enron's emails showed staff going around the org chart as the company failed.3

The org chart vs. the map in the inbox
Left: an org chart showing who reports to whom. Right: the same people, with lines drawn from their emails showing who actually works with whom, including many links that skip the chart. ORG CHART Who reports to whom. FROM THE EMAILS Who actually works with whom.
The orange lines are the knowledge an AI can't get from the org chart.

Every business has this record. Most owners haven't put theirs to work yet.

What it sells for

150+data sales through one shutdown service, SimpleClosure1
$10K-$100Ktypical payout per startup2
47%of SimpleClosure's shutdowns now include a code or data sale1
The big one: $10M for Spirit Airlines' records
Google paid 10 million dollars for Spirit Airlines' records: 500 million Teams messages, 100 million emails, and 176 thousand employee records. Teams chat messages 500M Emails 100M Employee records 176K
Google outbid an AI data company for it.1

The catch: these are private messages about real people who never agreed to become AI training material. Good sellers strip out personal details first.

A failed company's records sell for up to six figures. Yours are worth more, because you can still use them.

Put your records to work

A failed company's history gets sold once. Yours can work for you every day. It's exactly what an AI needs to work your way: how you price, who approves what, what went wrong last time.

Whether you own one business or a portfolio, that record is probably scattered across inboxes nobody can search. Organizing it is step one of AI that works. The market just put a price on it.

Ready to put your records to work? Book a free consult →

Sources
  1. "AI labs are buying data from dead startups." MIT Technology Review, Oct 13, 2026.
  2. "Failed Companies Are Selling Old Slack Chats and Email Archives to Train AI." Gizmodo, citing Forbes, Apr 17, 2026.
  3. Diesner, Frantz & Carley. "Communication Networks from the Enron Email Corpus." Computational and Mathematical Organization Theory, 2005.
  4. Yang et al. "The effects of remote work on collaboration among information workers." Nature Human Behaviour, 2021.
  5. Xu et al. "TheAgentCompany: Benchmarking LLM Agents on Consequential Real World Tasks." Carnegie Mellon University, arXiv:2412.14161.
§ · § · §