Case study
Case study: Satswana
What we built for Satswana, and what it has done so far.
Who
Satswana is a UK data protection company. It handles subject access requests and data protection work for its clients.
The problem
Subject access requests arrive with thousands of pages of email and documents. Every page has to be checked for other people's personal data, and that data found and redacted, inside a one-month deadline.
What we built
Review and redaction.
A pipeline that finds and redacts third-party data. A person checks every pack before it is released.
A knowledge base.
It indexes Satswana's email and documentation, over a hundred thousand emails and most of the company documentation, and drafts replies for a person to approve.
Secure delivery.
Response packs are delivered by link, with the password sent separately.
MinuteVault.
Meeting transcription and minutes that run end to end on Satswana's own machine, with nothing leaving it.
Satswana AI.
We designed Satswana AI, the platform Satswana runs on: a local open-weight model on Satswana's own hardware, with the vault family built on top (MinuteVault, MediaVault, RedactVault, KnowledgeVault, CRMVault and eVault). It is white-labelled, so it runs under Satswana's own name.
Since late June 2026
Results so far
70
subject access request cases
1,075
documents redacted
~186,000
redactions
About 186,000 redactions across 1,075 documents in 70 cases.
Our own count on 22 Oct 2026. A redaction is one redaction box, counted at word level.
“I am most impressed by James, because he has gone out there and learned about it, producing an excellent tool.”
James Howard is Head of AI at Satswana.