Olmec Dynamics
P
·14 min read

POPIA Compliant Automation: A Practical Guide to Zapier, Make and n8n

Is Zapier, Make or n8n POPIA compliant? Where your data goes, the POPIA sections that matter, cross-border rules and a 10-step checklist for SA teams.

Your sales team's new lead workflow runs beautifully. A form submission lands in Make, OpenAI scores it, the CRM gets a clean record and the right rep gets a WhatsApp alert. Then your information officer asks a simple question: where did that person's ID number, phone number and salary band actually go on the way?

For most South African businesses the honest answer is "several servers in Europe and the United States, plus a log we've never looked at." That isn't automatically a problem. POPIA does not ban cloud tools. But it does expect you to know where personal information travels, to have the right agreements in place, and to keep it secure. This guide shows ops and sales leads how to run POPIA compliant automation on Zapier, Make or n8n without stalling every project in a legal review.

This is practical guidance, not legal advice. For high-risk processing, have your information officer or attorney sign off.

The short answer: no tool is POPIA compliant on its own

You will see vendors and agencies describe a platform as "POPIA compliant". Treat that phrase with caution. Under POPIA, your business is the responsible party: you decide why and how personal information is processed. Zapier, Make and n8n Cloud act as operators, processing it on your behalf. Compliance comes from how you set up and run the workflow, not from the logo on the tool.

So the useful question isn't "is Zapier POPIA compliant?" It's "can we use Zapier for this workflow, with this data, in a way that meets our obligations?" Most of the time the answer is yes, once you've done four things:

  1. Mapped where the data flows and where each step is hosted.
  2. Accepted each vendor's data processing agreement and kept a copy.
  3. Minimised what passes through the automation and how long it's kept.
  4. Set up access controls and a plan for when something goes wrong.

Where personal information goes inside an automation

People tend to picture an automation as a pipe from App A to App B. In reality, a single run can copy personal information into five or six places:

  • The trigger app, such as your website form, Typeform, Gmail or a lead ads platform.
  • The automation platform itself, which receives the full payload of every run.
  • Execution history and logs, which store the input and output of each step so you can debug failures. This is the copy most teams forget.
  • AI steps, which send text or documents to a model provider that may be hosted somewhere different from your automation platform.
  • Intermediate storage, such as a Google Sheet, Airtable base or data store used as a lookup table.
  • The destination, such as your CRM, accounting package or a Slack channel.

Every one of those is a place where POPIA's conditions apply. If you're just starting out, our guide to choosing which business processes to automate first is a good way to pick a low-risk first workflow before you tackle anything involving sensitive records.

The POPIA sections that matter most for automation

POPIA has eight conditions for lawful processing, and all of them apply to automated workflows. In day-to-day automation work, these sections come up again and again.

Section 72: sending personal information outside South Africa

Almost every popular automation platform runs outside South Africa, so using one is a cross-border transfer. Section 72 allows this when at least one of five grounds applies. The recipient must be bound by a law, binding corporate rules or a binding agreement that gives an adequate level of protection, or the data subject consents, or the transfer is necessary for a contract with the data subject, or for a contract in their interest, or it's for their benefit when consent isn't practical.

In practice, most businesses rely on the first ground: a binding agreement. For a SaaS tool, that agreement is the vendor's data processing agreement (DPA). The adequacy test also covers onward transfers, so the agreement should restrict what the vendor does when it passes your data to its own subprocessors.

One trap to watch: special personal information (such as health, religious beliefs, race, biometrics or criminal behaviour) and children's information. Transferring it to a country without adequate protection needs the Information Regulator's prior authorisation under section 57(1)(d). If a workflow touches medical aid details or ID document scans, slow down and get advice.

Section 19: security safeguards

Section 19 requires "appropriate, reasonable technical and organisational measures" to protect personal information. You must identify foreseeable risks, put safeguards in place, check that they work, and update them as risks change. For automation, that means things like multi-factor authentication on the platform, least-privilege connections, restricted access to execution logs and a sensible retention period.

Section 21: a written contract with every operator

Section 21 says you need a written contract with each operator to make sure it maintains those security measures, and the operator must tell you immediately if it suspects a breach. Accepting the vendor's DPA is usually how you meet this for SaaS tools. If you use an agency or freelancer to build and maintain your workflows, they may also be processing personal information on your behalf, so your agreement with them should cover this too.

Section 22: telling people when something goes wrong

If there are reasonable grounds to believe personal information has been accessed by an unauthorised person, section 22 requires you to notify the Regulator and the affected people as soon as reasonably possible. Since 1 April 2025, the Information Regulator has required organisations to report security compromises through its eServices portal rather than by email. A leaked API key or a misconfigured webhook in an automation counts, so your incident plan should name who checks the logs and who files the report.

The stakes are real. Under section 109, the Regulator can issue infringement notices with administrative fines of up to R10 million, and it has done so, including a R5 million fine against the Department of Basic Education.

Zapier, Make and n8n: where your data actually sits

Here is what each vendor publishes about hosting, agreements and retention. Check the linked pages before you rely on them, because vendors update these terms.

Zapier

  • Hosting: Zapier hosts customer data on AWS servers in the United States and does not offer an EU-only or local storage option.
  • Agreement: Zapier publishes a standard Data Processing Addendum, which authorises international transfers and includes standard contractual clauses for EU data. It won't sign your company's own DPA.
  • Retention: according to its data retention page, content passing through Zaps is kept for 7 days in logs, 29 to 69 days in your account and up to 4 months in backups. Company and Enterprise plans can set a shorter account retention period.
  • Best fit: simple, low-sensitivity workflows such as lead notifications, task creation and internal alerts, where data leaving the country under a DPA is acceptable.

Make

  • Hosting: Make lets each organisation choose a US or EU data centre for storing and processing its data. There is no South African option.
  • AI caveat: choosing an EU zone doesn't guarantee every request stays in the EU. Make's own AI providers and data zones page notes that some AI features and all OpenAI models in Make's AI Provider route to the US, and bringing your own AI connection removes the EU guarantee.
  • Agreement: Make publishes a data processing agreement covering processing on your instructions, subprocessors and breach notification.
  • Retention: execution history retention depends on your plan, and Make lets you mark scenario data as confidential so it isn't stored in the logs. Check your plan's settings.
  • Best fit: complex multi-step workflows with branching logic, especially when you want an EU data centre and fine control over what each module stores.

n8n

  • Hosting (Cloud): n8n Cloud runs on Microsoft Azure with data hosted in the EU, and its DPA includes standard contractual clauses.
  • Hosting (self-hosted): you can run n8n on your own server, including one in a South African data centre. n8n's documentation is clear that for self-hosted installs it is neither controller nor processor, because it doesn't manage your data. The data stays where you put it.
  • AI caveat: n8n's subprocessor list shows AI providers processing in the US. And if your self-hosted workflow calls an overseas API, that step is still a cross-border transfer.
  • Best fit: workflows handling sensitive records, clients with data residency clauses in their contracts, and high-volume workflows where per-task pricing gets expensive.

How to choose: a quick decision guide

Work through these questions for each workflow rather than picking one tool for the whole business.

Does the workflow handle special personal information or children's data? If yes, keep it local if you can (self-hosted n8n in South Africa), minimise what passes through, and get advice before sending it offshore.

Does a client contract require data to stay in South Africa? Then cloud Zapier and Make are out for that workflow, because neither offers local hosting. Self-hosted n8n is the practical route.

Is it ordinary contact and business data, such as names, work emails and deal stages? Any of the three can work, provided you've accepted the DPA, recorded the transfer and tightened retention.

Does it include an AI step? Check where that model provider processes data, separately from where your automation platform is hosted. Strip out anything the model doesn't need before the call.

Who will maintain it? Self-hosting puts patching, backups and uptime on your side. If nobody owns that, a well-configured cloud tool is often the safer choice, because section 19 is about security as much as location.

Many businesses end up with a mix: Zapier or Make for everyday alerts and CRM updates, and a self-hosted workflow for the handful of processes that touch sensitive records. We help teams connect their whole stack this way so each workflow runs on the right platform.

A POPIA checklist for every automated workflow

Use this before a workflow goes live, and again when you change it.

  1. Write down the data. List each field the workflow touches and flag anything that's special personal information, account numbers or ID numbers.
  2. Map every hop. For each step, note the app, the vendor and the hosting region, including AI providers and storage you use along the way.
  3. Accept the DPAs. Accept or sign each vendor's data processing agreement, save a copy, and record which section 72 ground you're relying on.
  4. Send less. Map only the fields the destination needs. Passing an entire form payload "just in case" multiplies your risk.
  5. Shorten retention. Set log retention as low as your debugging needs allow, use confidential or no-log settings for sensitive steps, and delete test data.
  6. Lock down access. Turn on multi-factor authentication, limit who can view execution history, and connect apps with dedicated accounts rather than a director's personal login.
  7. Protect credentials. Store API keys in the platform's credential manager, never in plain text fields or shared sheets, and rotate them when someone leaves.
  8. Tell people. Update your privacy policy to say that data is processed outside South Africa and on what basis. For marketing messages, make sure you have the consent POPIA requires.
  9. Plan for a breach. Decide who gets alerted when a workflow fails or a key leaks, who checks the logs, and who reports to the Regulator through the eServices portal.
  10. Review it. Add each workflow to your records and review it at least once a year, or when a vendor changes its terms.

Example: an automation that handles financial documents

Some workflows deserve more care than others. We built a home loan PDF extraction workflow for a client whose team spent hours each week copying data from home loan PDFs into a Google Sheet. A web app takes the upload, Make extracts the data, OpenAI structures it and the result lands in Sheets.

Documents like these can carry ID numbers, income and banking details, so a workflow of this kind is exactly where the checklist earns its keep. The questions to settle up front are:

  • Which Make data centre does the organisation use, and where does the AI step process the document?
  • Does the extraction pull only the fields the team needs, or every page of the PDF?
  • Are execution logs set so full documents aren't stored after the run?
  • Who can open the Google Sheet, and is access reviewed when staff change?

The same thinking applies to our property valuation report app, which turns three PDFs into a bank-ready report. Automating document work saves hours, but only if the data handling is designed in from the start rather than bolted on after an audit.

If you're automating customer conversations rather than documents, our guide to WhatsApp lead follow-up automation covers the consent side of POPIA for sales messages.

Common mistakes we see

  • Forgetting the logs. The CRM is locked down, but the automation's execution history holds months of full payloads that anyone on the account can open.
  • Connecting with a personal account. If the workflow runs on a director's Gmail or a former employee's login, you lose control of access and of the audit trail.
  • Using test data from real customers. Test runs can store real records indefinitely. Zapier, for example, keeps test content until you delete the Zap.
  • Treating "EU hosted" as the end of the story. An EU data centre helps, but an AI step or a third-party app can still send data to the US.
  • Building it and walking away. Vendors change subprocessors and terms. A workflow that was fine at launch needs a periodic review.

Frequently asked questions

Is Zapier POPIA compliant?

Zapier can be used in a POPIA compliant way, but no tool is compliant on its own. Zapier stores data in the United States, so using it is a cross-border transfer under section 72. Most businesses rely on Zapier's data processing addendum as the binding agreement, then minimise data and shorten retention. For special personal information, get advice first.

Is Make.com POPIA compliant?

The same principle applies. Make lets you choose a US or EU data centre and publishes a data processing agreement. Choose the region deliberately, check where any AI modules process data, and use Make's confidential data setting for sensitive steps.

Does POPIA require data to stay in South Africa?

No. POPIA allows transfers outside South Africa when one of the section 72 grounds applies, most commonly a binding agreement that provides adequate protection. Local hosting becomes necessary mainly when a contract requires it, or when you're dealing with special personal information or children's information going to a country without adequate protection.

Is self-hosted n8n automatically POPIA compliant?

No. Self-hosting in South Africa avoids the cross-border transfer for the platform itself, but you take on the security obligations under section 19: patching, backups, access control and monitoring. Any step that calls an overseas API is still a cross-border transfer.

Do we need consent from customers to automate their data?

Not always. Consent is one lawful basis, but processing can also be justified when it's needed to carry out a contract with the person or to protect a legitimate interest. Direct marketing by electronic communication has stricter consent rules. What you always need is transparency: your privacy policy should explain what you process and where it goes.

Who in our business is responsible for this?

Every business has an information officer under POPIA, by default the head of the organisation. In practice, ops or sales leads who own a workflow should complete the checklist above and share it with the information officer before go-live.

Build automations your information officer will sign off

POPIA shouldn't be the reason your team keeps copying data between systems by hand. With the right platform for each workflow, signed agreements, lean data mapping and short retention, automation can be safer than the spreadsheets and email attachments it replaces.

If you'd like a second pair of eyes on your current workflows, or you're planning one that handles sensitive records, tell us what you want to automate. We'll map where the data goes, recommend the right platform and build it with POPIA in mind from day one. Start automating with Olmec Dynamics.