Kleinhans Digital

AI that respects POPIA, your people and your data.

I help South African organisations review how they use AI, then put the right local models, workflows and websites in place so sensitive data stays under their control.

Portrait of Jason Kleinhans
Jason Kleinhans, Founder

Who runs it

One person accountable for the decisions and the build.

Kleinhans Digital is run by me, Jason Kleinhans, an AI and automation specialist based in Johannesburg. I review how organisations actually use AI, then build the safer version, kept small enough that the details stay in view: what data is involved, where it goes, who can see it, and what the system is allowed to do.

I am not a lawyer. I help teams prepare better questions, reduce avoidable risk, and turn approved decisions into working websites, automations and local-AI tools.

How it works

One accountable human, and an AI team that does the rest.

This is the model I run, and the one I build for you. One person stays accountable and makes the calls, while an AI team handles the repetitive work and never goes home. The accountable human can be you, on your own hardware, or me on your behalf. Nothing client-facing goes out without a person signing it off.

Owner You Decisions and approval
AI Sales Enquiries, proposals, your CRM
AI Delivery Builds, launches, deploys
AI Research Signal and POPIA changes
AI Maintenance Keeps your systems healthy
AI Marketing Content, on a schedule
AI Other Custom to your workflow

Each role plugs into the systems you already use, your CRM, accounting, email and files, rather than replacing them.

Your CRMAccountingEmailCalendarFilesWhatsApp

What gets done

Advice you can implement, and implementation that knows why it exists.

The front door

AI use reviews

A plain-English look at where AI is already being used, what information is exposed, and which controls to agree before the habit spreads. Almost no South African organisation has checked this yet.

from R4,500 Book a review →
POPIA

Built around South African data rules, not bolted on afterwards.

AI automation, on your hardware

Assistants and workflows that automate the repetitive work and run on models you control, so sensitive data stays private. Still rare in South Africa: automation built to run fully local.

from R28,000 scoped per build

Recurring websites

Affordable websites for South African small businesses, maintained monthly so the public presence does not become another neglected system.

from R3,000/month

Most engagements start with the 10-Day Pilot below. From there: ongoing Managed AI Infrastructure from R8,500/month, or a single managed workflow from R2,500/month. Not sure where to start? The Pilot is built for exactly that.

The right way to start

The 10-Day POPIA-safe AI Pilot.

One real workflow. Automated end-to-end. On your infrastructure. In ten working days. No transformation deck, one working result you can point to.

What you get

One business process automated end-to-end and integrated into the tools you already use, your CRM, accounting, WhatsApp, calendar and files. Full documentation. You own the build.

Why ten days

Long enough to build something real. Short enough that the risk is completely bounded: fixed cost, fixed deliverable, no open-ended scope.

What it leads to

Most clients continue on a managed retainer so it stays reliable and compliant. Some do not, and that is fine. Either way you keep a working automation.

R18,500 once-off, fixed scope Claim your Pilot slot →

Best for firms with a repetitive admin, intake, reporting or follow-up process that is currently manual.

How the work stays safe

On your infrastructure Runs on hardware or a tenant you control. Client data does not leave the building.
Human approval points Agents prepare and route. A person signs off on anything that matters.
Least-privilege access Scoped access and an access-control review, not blanket keys to everything.
Security testing available Authorised security testing of the deployment, for regulated work that needs it.

Who I work with

Built for organisations that cannot be careless with data.

SMEs handling sensitive data

Five to fifty staff, with client lists, financials, ID numbers or records you would not want sitting on a server in another country.

Professional practices

Legal, accounting, medical and advisory firms where confidentiality is the work, not a nice-to-have.

Teams told to adopt AI

Leadership wants the productivity, but nobody has checked what data is leaving, or where it ends up.

Anyone already using public AI

If staff are pasting real client information into public tools, the exposure is already there. Better to make it safe than to ban it.

How it works

From a clear data picture to a working, explainable system.

  1. 01

    Review

    I map how AI is used today, what data it touches, and where the exposure sits. You get a plain findings report.

  2. 02

    Decide

    I help you decide what belongs in AI, what stays on local models, and the boundaries around data, access and retention.

  3. 03

    Build

    A small working implementation: a controlled assistant, a safer document workflow, or a website that supports the same rules.

  4. 04

    Hand over

    Documented in language your team can use, so the system stays understood and maintained, not quietly abandoned.

Proof of approach

Start with the data map, then decide what belongs in AI.

The first useful answer is often not a model choice. It is a clear record of the information involved, the people who need it, the systems it touches, and the parts that should stay out of public tools.

From there, I can help shape a small working implementation: a reviewable assistant, a safer document workflow, or a website and automation path that supports the same controls.

A hand-drawn notebook page laying out a simple, explainable system
A useful system is one your team can explain.
  • Map the data before choosing tools.
  • Keep sensitive work on controlled local models where required.
  • Document the boundaries in language the team can use.

Under the hood

An architecture, not a single tool.

The same system I run my own practice on, and build for the clients who need it. Cloud models for judgement, local models for anything private, a memory layer every part reads from, and workflows that keep running without a person being awake.

Control

Driven by chat and voice from your phone, so you can run it while you are out.

Reasoning

Strong models for judgement, drafting and the decisions that need care.

Private execution

Local models on hardware you control. Sensitive data never leaves the building. This is the part most vendors cannot offer, and it is still rare in South Africa.

Memory

A private knowledge layer that every part reads from before it acts.

Orchestration

Deterministic workflows, so the routine runs on its own instead of asking a model every time.

Delivery

Fast websites and automations wired into the same controls.

The specific tools shift every few months as the field moves. The architecture, and knowing what belongs where, is the part that lasts. That is what I build, and what I can build on your own hardware.

Questions

The things people ask before we start.

Is my data safe?

The sensitive work runs on local models on hardware you control, so client information does not leave the building. That is the whole point of the approach.

Do I have to replace the tools we already use?

No. I fit AI and automation around what you already run, and only change what genuinely needs changing.

How is this different from ChatGPT Team or Copilot?

Those send your data to servers in other countries. I build automation that can run on your own hardware, which keeps it inside POPIA.

How long does it take?

A use review takes days. A first working build is usually a few weeks, not months.

What does it cost?

The review is a fixed fee. Builds are priced by scope once we know what you actually need, so there are no surprise monthly bills.

Are you a compliance lawyer?

No. I handle the technical side, how data moves and where it lives. The legal compliance sits with your own advisor, and I work alongside them.

Contact

A calm conversation before any build.

If your organisation is already using AI, or your website and internal process need to catch up, start with a short note about what data is involved and what you are trying to improve.

A quiet desk detail for a considered digital advisory conversation
Small enough to stay considered.