Years of internal operational mess are about to have a market value
Companies are about to discover that years of internal operational mess have a market value.
Frontier AI labs are paying serious money for company data: email, Slack, Google Drive, Salesforce and Workday CRM records, ERP activity, and the ordinary work product employees generate while running a business.
That tells us something about where AI development is going.
The internet taught models what people know and say. Enterprise data can teach them how work actually gets done.
Consider a sales process. The final CRM record is useful. The emails before it, the internal debate over pricing, the ignored follow-up, the manager stepping in, and the exception someone made to close the deal contain something harder to manufacture: judgment in context.
That creates a strange second-order effect.
Companies have spent years treating this data primarily as a compliance liability and storage cost. Now some of it has an economic value precisely because it captures thousands of decisions nobody bothered to formalize.
The uncomfortable part is that selling it may also help train systems capable of absorbing pieces of the work that created it.
So the decision gets more complicated than “what is our data worth?”
Management has to price the cash today against privacy risk, employee trust, competitive leakage, and the possibility that its own operating history helps commoditize parts of its future labor.