Independent AI governance, audit, architecture and training. Scoped and delivered by the people who wrote the framework.
Built from an incident register, not a framework import. The controls we bring to your estate are the ones we run on our own: runaway resource consumption, controls that pass while blind, and reporting that is confidently wrong.
Every engagement leaves the system measurably better than the report found it.
Controls that hold when nobody is watching, because that is when they are tested.
Failures that recover without a person, and say plainly what they did.
Each incident becomes a control. That is where our catalogue came from.
From “which use case should we even start with” to “it is in production, governed, and someone owns it”. Each stage produces the next stage’s proposal.
Scores candidates on business pain against technical complexity.
Trust AI Governance. A scored assessment against 57 controls.
Thirteen fixed weeks, four approval gates. A failed gate moves the date, not the gate.
The run phase. Onboarded within ten working days of go-live.
The portfolio operating model, and the board reporting pack under it.
TAG and TechSight 90 are the two anchors. Most engagements begin at stage 2.
The consulting is built on things we run ourselves. Both are available today.
Thirteen languages, including South African English and Afrikaans with code-switching. Speaker diarisation, meeting summaries and action items, export to DOCX and SRT. Pay in Rands, top up as you go, no monthly commitment.
Self-improving, self-securing, self-healing, self-learning. The incident register that produced our control catalogue comes from operating it in production, which is why the controls hold.
Articles written for boards and the people who report to them, each carrying the date it was published. The brief goes out every morning.
A company that sells AI customer service agents published the research showing people yell and swear to escape one. The complaint is not the machine. It is the missing exit.
Read the article →