Glossary
Our trade has its words; your business shouldn't need a dictionary to buy from us. Each term below gets a plain definition and a link to the page that earns it. If a word we use isn't here, that's our mistake — tell us.
The system
Our name for the system we build. A private AI knowledge base that has read what your business already has — files, emails, history — answers questions in plain English with sources, and does routine work with your sign-off. It's owned software: licensed to you, running on your hardware. The name is ours; the thing itself is described plainly on the AI implementation overview.
The everyday name for what a Substrate is: a brain for the business, owned like the building. Five properties define one — it's yours, it has read your business, every answer shows its source, it does work, and nothing goes out without a signature. The full definition: What is a private AI brain?
The phrase you'll increasingly see abroad for the same idea: a business's own private AI knowledge system. Different flag, same thing. We build them for South African operating businesses, under South African constraints.
The pieces of the system that do work for you, added one at a time: customer relationship management, marketing, social, outbound chasing, and packs for your industry. Every module obeys the same rule — it drafts, a person signs. The catalogue lives on the products page.
The system runs on infrastructure you own or accounts you control — your servers, your cloud accounts, or a machine at your premises. We build, document, and hand over the keys. We never hold your data. What that means when you stop paying: the FAQ.
An open standard that lets one AI system work with different AI models. Because your system speaks it, you can use Claude, GPT, Gemini — or change later — without rebuilding anything. When better models ship, the same system gets better; the part we build is the part that lasts.
The BlitzBox configuration that runs open AI models on the machine itself, so no outside AI service is involved and nothing leaves your premises at all. For the hardware and security detail: inside BlitzBox.
The offer
The free first step: plain questions about how your business actually runs. You watch the business get written down as you answer, you keep the write-up, and it prices the engagement exactly — the quote it produces doesn't move afterwards. Begin it here.
One painful, high-value workflow, captured and running in six weeks at a fixed price. Most engagements start here regardless of size — it's the fastest way to see the pattern on your own data. Stages and prices: pricing.
The full brain for one department or unit, with three to five of your systems connected, so the whole team can ask it in plain English. Fixed in the quote.
One brain across the entire business, with the full handover: every record, every runbook, and the valuation pack. The flagship stage.
One system across a group of companies, with shared rules, group compliance architecture, and machines on premises where required. Priced per mandate.
The monthly retainer after a system is live: new context as the business evolves, new model capabilities folded in, and scheduled audit milestones with independent technology partners — the third-party verification a buyer's auditor wants to see.
The same private AI brain delivered as a machine: commissioned by hand, plugged in at your premises, with your business data staying on the box. Hardware once, intelligence monthly. The product: BlitzBox; the costs: BlitzBox pricing.
The asset
The owned software and records that make your business run — the map of how it fits together, the write-ups of how work gets done, the decision history — in a form an auditor can treat as an asset of the business and a buyer's adviser can verify. The argument in full: the asset page and the essay Operational IP, Now Capital.
Everything you receive, itemised: the code, the map, the write-ups, the decision history, the records of what the system did. Not deliverables in a folder — live, maintained artefacts you own from day one.
What advisers call a business whose real knowledge lives in a few heads. Buyers price it, successors inherit it, and every resignation triggers it. Capturing the knowledge in an owned, askable system removes the reason. The essay: Key-person risk; the sale-day version: AI before you sell.
The safeguards
The engineering term for our doctrine: the system drafts everything and sends nothing. Anything a customer, supplier, bank or regulator will ever see crosses a signature from a person in your business — by construction, not by setting. Why that's a feature and not a speed limit: the essay.
The Protection of Personal Information Act. In one line: your business stays responsible for personal information wherever it goes — including wherever staff paste it. That's the case for AI that keeps the information on systems you control. The plain-English tour (not legal advice): POPIA and private AI.
The agreement POPIA expects when another party handles personal information on your behalf — covering security, data residency, retention and breach notification. We sign one per engagement; it's part of the paperwork, not an extra. Our own commitments: the privacy policy.
Next step
Twenty minutes of plain questions — no term on this page required — and you get the first workflow we'd prove, the stage that fits, and the fixed price.
Begin the diagnostic