We build AI for work that has to hold up.
Assessments a board can act on. Programmes a funder can verify. Both running on infrastructure that never has to leave the country.
The thread between them: evidence is generated as the work happens, rather than assembled afterwards — so every claim traces back to the thing that produced it, months later, in front of whoever is asking.
Three products. One idea underneath them.
Different buyers, different rooms, the same discipline — the evidence is a by-product of doing the work, not a document assembled before an inspection.
Assessment & advisory
Reads an enterprise estate — the documents, the configuration, the data — and holds a scored position where every finding traces to its source file. Four assessment modules on one engine.
For boards and their partners → LumenAI-powered training delivery
Ingests your course material and proposes the programme, runs the weekly delivery engine, gives every learner an AI mate grounded in your own corpus — and emits the evidence pack as a query instead of a scramble.
For providers, employers and funders → ArgonSovereign AI infrastructure
The layer both products run on. Shared in South Africa, dedicated to you, or inside your own environment. Nothing is routed anywhere you have not agreed to in writing.
For whoever signs the risk register →Viki is the platform. Four solutions run on it.
One brain, one method, one evidence trail — pointed at four different decisions. Platform-agnostic by design, with its deepest grounding where the estates are hardest: SAP.
Evidence, not opinion.
Every enterprise decision of this size is a bet. The only question is whether it is placed on evidence, or on the confidence of whoever is presenting.
Three questions a board must answer.
Do we move platform — and when?
Where the estate genuinely stands against the platform change, and what it would actually take.
The assessment →What is our position on AI?
Whether AI belongs in this business, in which processes, in what order and at what cost.
The assessment →Is the programme actually on track?
An independent read on a programme in flight, taken from its artefacts rather than its reporting.
The assessment →The method behind all three is stack-neutral. The depth behind it is deepest on SAP — the honest version of that is here.
Viki is AI.
It is not an assistant.
It is a consulting brain, and it works the way your best consultant works — reading everything first, forming a position, and holding the evidence for it.
A brain that already knows the work.
What the platform does — not claims about what it did for someone else.
Ask it anything. It answers from the corpus.
“We are currently in Phase 1 (Discovery & Evidence Collection), and the next milestone is completing this phase to transition into Phase 2 (Assessment & Scoring).” [c1]
“I am unsure. The provided context does not contain information regarding specific cost levers.”
It would rather say nothing than invent something. That refusal is the single most important thing about it. How the gate works →
Four modules. One brain.
One context, not four disconnected conversations. Every engagement makes the corpus deeper, so a second assessment starts from the first rather than repeating it.
Everyone reaches the same standard. Nobody has the same week.
Training is still built around a constraint that no longer exists — one room, one pace, one instructor talking at the speed of the median. Lumen takes the programme apart into work: dated targets per learner, a tutor each in the shape of Lumi, and support routed to whoever is quietly drowning rather than whoever asks loudest.
And because the evidence is written while the work happens, the compliance pack is a query instead of a scramble.
- 01Ingest what you already have
A qualification spec, your decks, your SOPs — parsed, indexed, and turned into a proposed programme a person approves.
- 02A tutor each, at eleven at night
Lumi answers in the context of the exact activity the learner is stuck on, from your corpus and nowhere else.
- 03It catches the quiet ones
A learner who is confused asks. A learner who is drowning goes quiet — and that is the signal a register cannot see.
- 04The proof writes itself
Every stage writes an audit event as it happens, so reporting is extracted rather than reconstructed.
If you are the client.
- 01Evidence you can audit
Every figure carries the source file it came from.
- 02Weeks, not quarters
One analyst and a brain, rather than a floor of consultants.
- 03Your data stays yours
Sovereign deployment is your call, not ours.
If you are the partner.
- 01Subcontract, or white-label
We deliver behind you, or under your brand.
- 02Your consultants, your badge
The deliverable comes out in your house format.
- 03Your margin in full
You price your client — that spread is yours.
- 04Non-exclusive, either way
A sequence, not a lock-in.
Worried about handing your business to someone else's AI?
That is exactly why Argon exists.
Both products run on Argon.
Assessment & advisory — the four modules
Programme delivery & assurance
Sovereign AI infrastructure — the layer both run on
AN AFFILIO PLATFORM
Sovereign
Isolated tenancy on our infrastructure, in South Africa
Reserved to you
Not shared, not scheduled against other engagements
Inside your boundary
On your floor — nothing leaves the building
Where you elect it
Declared in your engagement, never the default
Your data never has to leave your control. Argon →
Start with a conversation, not a sales pitch.
A 45-minute scoping session. We will tell you which assessment fits, what evidence it would need, and whether you need one at all.