How it works
Four steps from a business's archive to a licensed dataset.
We do the work in between: checking rights, protecting personal data, and turning many small collections into something an AI company can use.
01
Assess
We look at what data you hold and whether there is demand for it from AI buyers.
Every engagement starts with a conversation. We learn what the business does, what material it has produced over the years, and where that material lives today.
We then compare it with what AI companies are asking for. Some data is in demand on its own; some becomes valuable only when combined with data from similar businesses. If we don't think there's a market for it, we say so.
For the business
A short call and a rough description of your data.
For the buyer
Sourcing is driven by real demand, not speculative collection.
02
Verify rights
We confirm the business created or owns the data and has the right to license it.
We establish, source by source, whether the business created the material, commissioned it with full rights, or obtained it from someone else. Only material the business has the right to license goes forward.
We also identify any personal data — customers, staff, third parties — and decide with the business whether it will be anonymised or excluded. The conclusions are recorded in writing and signed off by the business.
For the business
Answers to our questions about where material came from, and a signed confirmation.
For the buyer
A provenance record for every source in the dataset.
03
Clean and combine
We anonymise, standardise and pool data from many similar businesses into usable datasets.
We remove or anonymise personal information, strip out duplicates and irrelevant material, and convert everything into a consistent structure and format.
Data from many similar businesses is then pooled into a single dataset, large and varied enough to be useful for training. Each record keeps a link to its source so provenance is never lost.
For the business
Nothing — this step is ours.
For the buyer
Standardised, deduplicated data with a stated anonymisation method.
04
License
We license the dataset to AI companies with full documentation and share the revenue.
We license the finished dataset to AI companies under clear written terms covering permitted uses, restrictions and duration.
Revenue from each license is shared with the businesses whose data is included, according to the terms they agreed at the start.
For the business
You receive your share when the dataset is licensed.
For the buyer
One agreement, one invoice, full documentation.