Skip to main content
Impart GlobalBegin the conversation
Disciplines // Order // POPIA-Safe AI

AI workflow automation that clears the compliance bar before it ships.

Automate the repetitive work your team should not be doing, on infrastructure that keeps personal information inside your control. Built from day one with data minimisation, audit logging, and lawful-basis mapping. Ready for Information Regulator scrutiny.

Carries
regulated-data deployment
Scale
tenant-private
01 // The shape of it

Most AI pipelines you will see pitched in South Africa send personal information to US-based LLMs with no data residency guarantees and no POPIA-aligned processor agreements. We build the harder version. Prompts are scrubbed before they leave your environment. Responses are logged with who, what, and why. Models run in-region where possible. Every workflow comes with the paperwork your information officer will want to see.

02 // What it gives you

What the work returns.

  1. 01Data Minimisation by DefaultPrompts are scrubbed of personal information before they reach any external model. You decide what leaves, and what stays inside your walls.
  2. 02Full Audit TrailEvery AI call is logged with prompt, response, user, time, and purpose. When your information officer asks what the model saw, you have a definitive answer.
  3. 03Processor Agreements ReadyWe supply POPIA-aligned operator agreements covering every third-party model in the pipeline, ready for your legal team to review.
  4. 04In-Region Where PossibleFor workloads that justify it, we run open-weight models in SA or EU regions instead of defaulting to US-hosted APIs.
  5. 05Lawful Basis MappedEvery workflow is tied to a documented lawful basis: consent, contract, legitimate interest, or legal obligation. No ambiguity.
  6. 06Fits Your Existing StackWe integrate with Microsoft 365, Google Workspace, Salesforce, HubSpot, and any line-of-business system you already run.
03 // How it runs

How the work runs, in order.

  1. Function Scoping

    Lock the target workflow, the data it touches, and the lawful basis for processing.

  2. Data Flow Mapping

    Every field, every system, every hop. Visualised and signed off by your information officer.

  3. Guardrail Design

    We spec the scrubbing, logging, retention, and exception-handling rules for the workflow.

  4. Build

    Engineering against the guardrail spec, with production-grade observability from commit one.

  5. Audit Pack

    Operator agreements, DPIA template, audit log configuration, and runbooks handed over with the system.

  6. Deployment

    Go-live with monitoring, plus a thirty-day review where we tune against real usage.

04 // What you keep

What stays when we step back.

  • An AI workflow handling the work your team currently does manually
  • POPIA-aligned operator agreements covering every model in the pipeline
  • Full audit logs tying every AI decision back to a lawful basis
  • A workflow your information officer can sign off with a clear conscience
  • Production monitoring that catches drift before it becomes a breach
05 // Questions

What people ask first.

Why does POPIA matter if my competitors are not worrying about it?

Because the Information Regulator is enforcing. Penalties can reach ten million rand or imprisonment, and class-action risk is live. Building POPIA fit in after the fact costs three to five times more than building it in from day one.

Can you use ChatGPT or Claude for workflows with customer data?

Yes, if the data is scrubbed, the processor agreements are in place, and retention is controlled. We do that work, so leading models can be used safely.

What data residency can you achieve?

For workflows that warrant it, we deploy open-weight models in SA or EU regions. For workflows that do not, we stay with US-hosted APIs under POPIA-aligned processor agreements. The choice depends on the data class and the cost model.

How long does a typical build take?

Six to twelve weeks from kick-off to production for a single workflow. Longer for multi-workflow programmes. We share a weekly working build from week two onwards.

What happens if POPIA changes?

We track regulator guidance and update deployed workflows as the rules evolve. Clients on our AI Operations retainer get those updates as part of the service.

07 // Begin

Make something with us.

Artists, builders, patrons and the curious: tell us what you're working on. We answer within one business day.

  1. i.You write. A few lines on what you are making or building.
  2. ii.We answer. Within one business day, from the person who would do the work.
  3. iii.We find the fit. A collaboration, a commission, a tool or a venture: where the work belongs.
Rather talk? Ask us anything on WhatsApp. (opens in a new tab)

Within one business day.