Buying private AI can sound like buying a data centre. It does not have to. For a founder or office manager, the useful starting point is much simpler: name a recurring task, delay or decision that consumes too much human attention. The Halo AI turns that practical problem into the brief, system and handover needed to run it privately.
The whole path
- Describe the work in ordinary language.
- Review a right-sized recommendation with a person.
- Receive a working private system, not a box of parts.
- Keep sensitive actions human-gated and the operating record visible.
01 · Ask — start with the problem.
Tell the public Halo AI what you want to improve, automate or protect. You might describe supplier paperwork that keeps stalling, customer cases that require information from four systems, or a daily queue that only one person knows how to clear. You do not need to know which AI model or server could handle it.
02 · Qualify — make the need concrete.
The Halo AI asks about the outcome, the people involved, the documents and systems behind the work, and which decisions must stay with a person. Together, these answers become a short opportunity brief: what useful result should change, how often the work happens and what boundaries matter.
03 · Specify — turn the brief into requirements.
Next, the Halo AI translates the approved brief into practical requirements: workload, users, document volume, connections, privacy needs and acceptance checks. You do not have to write an RFP from scratch. If your organisation already has one, it can be reviewed; if it does not, the buying brief is prepared for you.
04 · Procure — choose the right Box.
The recommendation matches the work to a model, capacity and operating mode, then shows the price range and room for growth. A person reviews the recommendation with you before purchase. The aim is not the biggest machine in the catalogue; it is a system that handles the first job and has sensible capacity for what you may add next.
05 · Deliver — receive a working boundary.
Your SovereigntyBox arrives with the approved configuration, local AI model, Halo AI and system record. The handover is visible: you can see what was prepared, what remains to be added and which outside connections are off, optional or approved. You are receiving an operating system for the work, not an assembly project.
06 · Continue — resume privately.
The project continues on the hardware you control. You choose which business documents and system connections the Halo AI may use. Those records are read and organised on your machine, and nothing moves outward silently. A fully offline installation can work only with the information and tools inside your environment.
07 · Prove — see what is running.
The Sovereignty Report shows the system boundary, where its components came from, who can reach it and what changed. The Halo AI also keeps sources attached to answers and records meaningful actions. That gives an owner, auditor or IT partner something more useful than a promise: a reviewable operating record.
08 · Expand — add the next useful Halo AI.
Once the first workflow is proven, you can add another Halo AI for compliance, commerce, operations, finance or customer work. Each role gets only its approved documents, tools and permissions. Routine work can move faster while judgement, money, customer relationships and sensitive actions still return to people.
Open the Halo AI.
Describe the first task or bottleneck in your own words. No server specification required.
Open the Halo AI