Home » Insights » When a Client Declines an AI Tool: Your Non-AI Pathway Explained

When a Client Declines an AI Tool: Your Non-AI Pathway Explained

Client declines psychology AI tool

One of the most specific obligations in the updated HPCSA Booklet 20 is easy to overlook in the broader conversation about AI compliance: when a client declines an AI tool, they may not be disadvantaged or refused access to your services as a result (Health Professions Council of South Africa, 2025). This is not a general principle — it is an explicit requirement. And it has a direct practical implication that most practitioners have not yet fully worked through: you must have a functional, ready-to-use non-AI pathway for every clinical function in which AI currently plays a role. You can read a full overview of what Booklet 20 requires in our guide to AI and digital ethics in psychology practice.

This post focuses specifically on what that pathway must look like — how to have the consent conversation, what your non-AI alternatives need to cover, and how to document the client’s decision without creating a two-tier service.

Why the non-AI pathway requirement matters

The requirement for a non-AI pathway reflects a core principle running through Booklet 20: the standard of care must remain consistent regardless of whether AI is involved (Health Professions Council of South Africa, 2025). A client who declines AI documentation tools must receive the same quality of clinical notes as a client who consents. A client who declines an AI-assisted assessment platform must receive the same quality of assessment as a client who does not.

Consequently, “we only use AI tools” is not a compliant response to a client’s refusal. Neither is offering the client reduced services, longer waiting times, or a lower standard of documentation as the default outcome of declining. Furthermore, the obligation applies across all AI-assisted functions in your practice — not only the most visible ones. If your telehealth platform uses automated mood tracking, your assessment tool uses algorithmic scoring, or your session documentation relies on a cloud-based scribe, each of those functions requires a non-AI alternative.

For guidance on which of your tools may require evaluation before use, see our article on checking your AI tools for POPIA compliance.

Having the informed consent conversation

The consent conversation about AI use is not the same as a general digital consent discussion. HPCSA Booklet 20 requires that clients understand what the AI tool does, what data it accesses, how that data is stored, and what their right to decline means in practice — before the tool is used (Health Professions Council of South Africa, 2025). This conversation needs to be both clinically appropriate and culturally accessible.

In practical terms, this means using plain language rather than technical descriptions. You do not need to explain how a transcription algorithm works — you need to explain that a third-party service records and summarises the session, that the recording is stored on an external server, and that the client can say no without any impact on their care. For clients with limited digital literacy, or for whom privacy concerns carry additional weight due to community, cultural, or occupational context, this conversation may need to be more detailed and given more time.

Specifically, your intake process should include a dedicated AI disclosure clause that covers each tool in use — not a single catch-all reference to technology. The clause should name the category of tool (for example, AI session summarisation), describe the data involved, and provide a clear opt-out. Where a client declines before the first session, this should be noted in the clinical record before any AI tool is activated in their care.

What your non-AI documentation pathway must include

For most psychology practices, the most immediate non-AI pathway to establish is for session documentation. If your current workflow relies on an AI scribe or summarisation tool, your non-AI alternative must produce a clinical record of equivalent quality — not a shorter note, not a different format, not a delayed note. The standard of documentation must be consistent (Health Professions Council of South Africa, 2025).

This does not mean handwriting notes during sessions — that is rarely consistent with therapeutic presence. It means having a clear manual process ready: a note structure you use, a time allocation you protect, and a template that captures the same clinical information your AI-assisted notes would contain. For practitioners who have relied heavily on AI documentation tools, rebuilding this manual workflow before a client declines is far easier than scrambling to establish it after the fact.

Additionally, if you use AI tools for other functions — risk assessment flagging, mood tracking integration, assessment scoring — each one needs a manual equivalent. Document what that equivalent is and how it is accessed. This is also good governance practice for when tools become unavailable or change their terms of service. For a step-by-step guide to reviewing AI session note workflows for compliance, see our article on how to document AI session notes compliantly.

Documenting the client’s decision

When a client declines an AI tool, that decision should be recorded in the clinical file — briefly and factually. A single sentence noting that the client was informed of the use of AI in a specific function, that they declined, and that a non-AI alternative is in use is sufficient. This protects both the client and the practitioner if the matter is ever reviewed.

What should not appear in the record is any language that frames the client’s decision negatively — as non-compliance, as a complication, or as a limitation on the service provided. The client is exercising a right explicitly recognised by the HPCSA. The record should reflect that the right was offered and the appropriate pathway followed, nothing more.

Furthermore, if a client initially consents to AI use and later withdraws that consent, the withdrawal applies from that point forward. Any AI-processed records from before the withdrawal remain in the file — you cannot retroactively remove them — but no further AI processing of that client’s information should occur after the withdrawal is received.

What about group practices and supervision contexts?

The non-AI pathway obligation applies at the practice level, not only to individual practitioners. If you manage a group practice or supervise other psychologists, you are responsible for ensuring that non-AI pathways exist across the team — not only for your own caseload (Health Professions Council of South Africa, 2025).

In practice this means your practice’s AI use policy — which Booklet 20 expects all practices to have — should include a section on non-AI pathways: which functions have AI alternatives, what those alternatives are, and how they are accessed. New practitioners joining the practice should be oriented to both the AI tools in use and the non-AI equivalents from day one. This is particularly relevant in supervision contexts where interns or registrars may have had limited exposure to manual documentation workflows.

Building this kind of governance structure around AI is not only a compliance requirement — it also protects the quality of care when tools fail, change, or are withdrawn. A practice that depends entirely on a single AI platform for session documentation has a fragile workflow regardless of the consent question.

Staying current with the HPCSA’s evolving AI standards is exactly what CPD is designed to support. Explore the PsyCampus 2026 CPD packages — HPCSA-accredited, relevant to current practice, and available across all registration categories — or check whether you qualify for a recent graduate discount of up to 40%.

References

Scroll to Top