Blog

Why AI companies buy everyday work records

· 4 min read

A good dispatcher knows what to do when a shipment gets stuck at the border. An experienced quality inspector spots a defect before it reaches the customer. This knowledge is not in any textbook, but it is written down in a company's everyday records. That is exactly why AI companies pay for such records.

An open filing cabinet drawer full of document folders

Why the internet is not enough

AI models have learned mostly from publicly available text: articles, books and forums. There you can read how work is supposed to be done. How it is actually done rarely shows up there: how a late shipment is re-planned, how a customer complaint is resolved, how a defect is found and fixed.

AI companies are now building systems that are meant to help with exactly this kind of work. Training them takes real examples, and those sit in companies' internal systems, not on the internet.

Which records can be valuable

This is not only about IT companies. Records from any industry can be valuable if they show how people handled everyday work situations:

  • Logistics and transport: dispatch records, delay and incident notes, correspondence with customers and carriers.
  • Manufacturing: quality control and defect records, engineering change notes, equipment maintenance logs.
  • Construction: project correspondence, defect reports and change orders.
  • Retail and services: customer complaints and the replies to them, support tickets.
  • IT: source code with its change history, tasks, bug reports and documentation.

All of these have one thing in common: they show a problem, what a person did about it, and the result. AI learns best from examples like these.

What sets the value

Not every archive is equally valuable. The price depends most on:

  • Length of history: how many years of work records have been kept.
  • Volume: how many people worked at the company and how much they wrote down.
  • Format: whether the records are digital, for example in e-mail or company systems, and not only on paper.
  • Completeness: whether the records are connected, for example a complaint together with its reply, or a defect together with how it was fixed.

If the company was one person or worked mainly on paper, there is probably no data to sell. You can get a rough estimate with our data value calculator.

What is not sold

AI companies need examples of work, not information about particular people. So before the data reaches a buyer:

  • names, contact details, personal ID numbers and other personal information are removed;
  • HR files, salaries, health data and passwords are not taken at all;
  • the buyer's licence forbids any attempt to re-identify people or companies.

More on this in our article on how anonymisation of company data works.

How it works

  1. You fill in a short questionnaire about the company and its data: what the company did, how long it operated and which systems it used. No data is handed over at this stage.
  2. We look for a buyer. The buyer sees only a one-page description with no company name and no data.
  3. Once a buyer is found, the data is copied to secure storage in the European Union and anonymised. A data processing agreement (GDPR Article 28) is signed before anything is copied.
  4. The buyer receives the anonymised dataset, and the money goes to the insolvency estate or the company.

There is no upfront cost. If the data isn't sold, you pay nothing.

Experience that can stay useful

Behind every archive are years of people's working experience. When a company closes, that experience usually disappears along with the accounts and the computers. It can stay useful: anonymised, with no names or personal data, and with a return for creditors or owners.

If you have a company in insolvency or liquidation, write to us before anything is deleted. On what can be sold, see our article can an insolvency administrator sell a company's data?

This article is general information, not legal advice.