01 / FIT — TEAM
Workshop
Half-day · on-site or remote
Your team builds one working AI workflow with me. Walks out with a prototype, not a slide deck.
Speak to Paul →
Live · Operator available
Production AI
without an
engineering team.
DomanskiAI installs the methodology that built MzansiEdge solo — sports-betting intelligence platform, live in market, no engineers hired. Workshop, Diagnostic, or Retainer — pick the shape that fits the gap.
// Headcount Zero · v1.0
§ 01 · The wedge
You can hire an advisor who recommends what to build.
You can hire an advisor who recommends what to build.
Or an operator who builds it.
Most people selling AI in this market have never shipped a production system. They package OpenAI documentation as expertise and bill by the deck. I ship.
// COL.A Credentialed advisor
- Recommends what to build
- Bills for the plan
- Hands off to your engineers
- Optimises for the next engagement
- Sells time
// COL.B Operator (DomanskiAI)
- Builds it
- Bills for the system
- Replaces the need for them
- Optimises for the kill switch
- Sells output
§ 02 · Engagement shapes
Three shapes. Pick the one that fits.
Half-day / 3 weeks / monthly. All conversations start the same way.
02 / FIT — PRODUCT
Diagnostic
Three weeks · scoped delivery
I deploy one production AI system in your environment, with a runbook for your team.
Speak to Paul →03 / FIT — SURFACE
Retainer
Monthly · continuous
I am the engineering team for one product surface.
Speak to Paul →
§ 03 · The methodology
Read the full methodology →
Headcount Zero
Four layers. Break one, break the system. Headcount Zero treats AI agents as the labour layer of a software business and the human operator as the technical director.
01
DIRECTION
One human. Sets scope, reviews output, kills bad work.
Doesn't write code line-by-line; does write the briefs. The only human in the loop.
02
ORCHESTRATION
Every unit of work is a brief.
Acceptance criteria · dependencies · evidence required to close. Briefs are the contract that scales.
03
LABOUR
Multi-agent stacks. Parallel where independent. Sequential where dependent.
Currently Claude Code. The methodology survives whichever model is current.
04
EVIDENCE
Gates verify before anything ships.
The test passing is evidence. The agent saying it works is not.
§ 04 · Proof
Two products built solo. One operator. The methodology is the credential.
LIVE IN MARKET
// product 01
SPORTS · INTELLIGENCE
MzansiEdge
Sports-betting intelligence platform. Real-time evidence pipelines, automated narrative generation, full publishing automation across web, Telegram, and email. Multi-agent Claude Code stack. Solo operated.
EVIDENCE PACK
NARRATIVE GEN
VERDICT VALIDATOR
CHANNEL PUBLISH
ROLES REPLACED → ~5–6 FTEs · COST → token spend + 1 director
// PRODUCT 02
PAUSED
AdFurnace
Programmatic ad operations engine. Same architectural pattern. Paused while bandwidth is directed at DomanskiAI.
// CASE STUDY 01
SOON · 19 MAY
Mack Brands
First DomanskiAI engagement. Half-day Workshop. Distributor pitch deck workflow for a premium spirits brand house.
Read the case study →
The pattern repeating across two unrelated domains is the methodology proof — not whether both are live this week.
§ 05 · The operator
Who I am.
Paul Domanski. Cape Town. Solo operator running DomanskiAI.
No degree. No corporate CV. No alumni network.
Two AI products built and operated alone. The methodology that built them — Headcount Zero — is what DomanskiAI sells.
If a buyer asks "have you done this before," the answer is: MzansiEdge is running right now. Pick a screen-share slot and I'll log in.
If your roadmap is slipping and your hiring pipeline is dry —
that's the conversation I want to have.
30 minutes. Free. Qualifying only. By minute 25 we both know if there's a fit.