SAHPRA’s Phase I Trial Guidelines Are Now in Effect — Here’s How Small SA Pharma Teams Handle the Extra Paperwork

Your regulatory affairs lead has just finished the first read-through of SAHPRA’s Pre-Clinical Testing Guidelines and Clinical Requirements for Phase I and First-in-Human Studies. They took effect on 1 June 2026, and they add a new layer of documentation to every Phase I application your team puts together from here on. That’s stacked on top of SAHPRA’s existing job: reviewing every clinical trial application for an unregistered medicine, and any new indication or dosage for a medicine that’s already registered, under the General Regulations Made under the Relevant South African Acts.

More sections to draft. More cross-checking against requirements that didn’t exist six months ago. Same five-person team, same budget, and the business case for another regulatory affairs hire still hasn’t made it past the finance meeting.

So what do you actually do with the extra load? The answer that’s spreading across regulated industries — and one SA’s own tax authority is already using — isn’t to hire. It’s to change who writes the first draft.

The Opportunity

Here’s the number that should change how you think about this. McKinsey’s analysis of pharmaceutical regulatory workflows found that AI tools can take an existing protocol and produce an “80%-ready” first draft of the documents that follow from it — in minutes rather than days — cutting overall document creation time by close to half. IntuitionLabs reports something similar from the clinical safety side: on specialised AI platforms, safety narratives that normally take days to write come together in minutes, around 90% faster than the standard process.

Both of those figures are global. Nobody’s published an SA-specific version yet, and we’re not going to pretend one exists. But the direction lines up with everything else in the research on this topic: AI produces a first draft, a qualified person checks it against the rules that actually apply, and only the checked version goes anywhere near a submission.

That’s the reframe worth sitting with. This isn’t “AI replaces your regulatory affairs person” — your team still needs someone who knows SAHPRA’s requirements and the GRMRSA inside out. What changes is where that person’s hours go. Less time staring at a blank page, more time on the part of the job that was always theirs to begin with: checking that what’s written is actually right. And that’s exactly the part of the workload the new Phase I/FIH guidelines just made bigger.

For a five-person team, that distinction is the difference between absorbing the new requirements with the people you’ve got, and needing a sixth person just to keep the submission queue moving. Nobody’s promising the workload disappears. But if the hours that used to go into writing now go into checking, the same headcount can cover more ground than it could in May 2026.

How It Works

Strip away the branding and the workflow is straightforward. An AI drafting tool takes what you already have — the trial protocol, previous submissions, source documents — and produces a first-cut version of whatever comes next: a protocol section, structured submission content, clinical study report text, or a mapping of a new guideline against your existing controls. Your regulatory affairs or QA reviewer then goes through that draft against SAHPRA’s requirements and the GRMRSA, checking every reference and every claim. They sign off. Only then does the document go into the submission package.

McKinsey names four places this shows up in pharma regulatory work: automated structured content authoring, automatic updates to source documents when something upstream changes, automated translation, and generative drafting of clinical study report content. None of those four remove the reviewer from the process. What they remove is the hours spent getting from a blank page to something worth reviewing.

There’s a regulatory reason this division of labour matters, beyond saving time. EY’s research on AI in GxP environments is direct about it: AI tools used for compliance or validation work need subject matter expert review and human oversight before their output counts as acceptable evidence. That review step isn’t an extra layer bolted on for comfort — it’s the thing that makes the AI-assisted draft usable at all.

You’ve got a local precedent for exactly this pattern. SARS has been moving toward AI-driven predictive compliance monitoring — flagging likely problems before they turn into missed deadlines and Administrative Penalties, with a person still making the final call on every flag. Swap “VAT deadline” for “SAHPRA submission date” and “tax flag” for “regulatory document draft,” and it’s the same structure: AI does the first pass, a person decides what happens next. If your team already trusts that pattern for SARS, there’s no real reason to trust it less for SAHPRA.

Case Study: A Five-Person Regulatory Team’s First Submission Under the New Guidelines

This scenario is illustrative — we haven’t found a published South African case study on this yet, so treat the company below as a composite rather than a real business.

A five-person regulatory affairs team at a small Cape Town biotech was partway through preparing a Phase I/FIH application when SAHPRA’s new guidelines took effect on 1 June 2026. The new requirements added several sections to a submission they’d already started drafting, on top of the existing GRMRSA review process. There was no budget for a sixth team member, and the submission deadline didn’t move.

What changed: the team set up an “AI drafts, regulatory lead authorises” process for the new sections. Their AI tool worked from the existing protocol and submission templates to produce first-cut drafts of the additional content the new guidelines required. The regulatory affairs lead then went through each draft against the current SAHPRA and GRMRSA requirements before anything was finalised — checking structure, checking references, checking that nothing had been left generic where it needed to be specific.

The result, in line with the global benchmarks above (though not measured for this team specifically): the time spent getting from nothing to a workable draft dropped sharply, and the lead’s hours shifted from writing the new sections from scratch to checking what the AI had already produced.

Here’s the friction, and it’s the part that matters most. On the first pass, the AI’s draft cited an older, more general regulatory reference instead of the new Phase I/FIH-specific requirement — the exact update the team was trying to account for in the first place. It read smoothly. It would have passed a quick skim without anyone noticing. It only got caught because the sign-off step meant checking each citation against the current guidelines line by line, not just reading the draft for tone and structure.

That’s the entire argument for why human authorisation isn’t optional here. A draft that looks finished and a draft that’s actually compliant with the 1 June 2026 requirements aren’t the same thing — and the gap between them is exactly where the regulatory affairs lead’s experience does its work.

Frequently Asked Questions

We’re already paying for accounting software and an ERP system — can we justify another subscription just for drafting documents?

We don’t have confirmed ZAR pricing for AI document-drafting tools aimed at SA pharma SMBs, so we’re not going to make up a number. But turn the question around: what does it cost you not to do this? SARS issues Administrative Penalties for missed deadlines because non-compliance has a real price, and the same logic applies to a delayed SAHPRA submission — every week a Phase I application sits unfinished is a week added to your timeline to trial. The realistic alternative to a subscription isn’t “no cost.” It’s either a delayed submission or another regulatory affairs salary. Against either of those, a drafting tool is the cheaper option, even before anyone confirms the exact figure.

Will SAHPRA actually accept a submission that started life as an AI draft?

SAHPRA hasn’t published a position on AI-assisted drafting, in either direction, and we won’t pretend otherwise. What we do know, from EY’s research on AI in GxP environments, is that human sign-off — which SAHPRA submissions already require no matter how the first draft was produced — is what makes AI-assisted output acceptable as compliance evidence. The document SAHPRA receives is the one your regulatory affairs lead has checked, corrected where needed, and approved. Where the first draft came from doesn’t change what SAHPRA is reviewing.

Can we put protocol details or patient data into an AI tool without running into POPIA problems?

We haven’t found SA-specific POPIA guidance covering clinical trial data in AI drafting tools, so this is a place to be careful rather than assume it’s fine. The practical approach: keep AI use to drafting and structuring de-identified protocol and regulatory text — the parts of the document that don’t contain anything patient-identifiable. Your regulatory affairs lead stays responsible for deciding what identifiable data, if any, ends up in the final submission. If you wouldn’t email it to an external contractor, don’t paste it into the AI tool either.