AgenticSA AI Acceptable Use & Responsible AI Policy

AgenticSA believes AI should help businesses increase capacity, improve customer experiences and reduce repetitive work while maintaining appropriate human responsibility.

Our approach is based on:

AI assists. Humans remain accountable.

AgenticSA is designed to support responsible human + AI collaboration rather than the uncontrolled replacement of human judgement.

This Policy applies to AgenticSA functionality including:

• SMS AI Agents;
• WhatsApp AI Agents;
• Email AI Agents;
• AI Voice Agents;
• AI Front Desk Employees;
• AI-generated responses;
• AI-assisted workflows;
• lead qualification;
• appointment assistance;
• customer-service automation;
• AI-generated summaries;
• conversation analysis; and
• future AgenticSA AI functionality.

Customers determine how their AgenticSA AI Employees are deployed within their businesses.

Customers remain responsible for:

• selecting appropriate use cases;
• supplying accurate training information;
• establishing lawful authority for processing information;
• obtaining required communication permissions;
• monitoring AI performance;
• maintaining appropriate human oversight;
• correcting inaccurate business information;
• defining escalation procedures;
• honouring opt-outs;
• complying with applicable professional obligations; and
• determining whether AI is appropriate for a particular decision or interaction.

AgenticSA's technical ability to perform an action does not necessarily mean that the action is legally or professionally appropriate for every customer.

Customers must not deliberately configure AgenticSA to deceive individuals into believing an AI system is a human where such representation would be misleading.

Where required by law, provider rules or the nature of the interaction, appropriate disclosure must be made that an individual is interacting with AI.

For Voice AI in particular, AgenticSA recommends clear disclosure at the beginning of appropriate AI-powered calls.

AI Employees should have clearly defined boundaries.

Customers should configure escalation to a human where appropriate, including for:

• serious complaints;
• emergencies;
• unusual circumstances;
• disputes;
• sensitive personal matters;
• high-value transactions;
• professional advice;
• pricing exceptions;
• refund disputes;
• legal threats;
• complex customer problems; and
• matters outside the AI Employee's approved knowledge or authority.

Artificial intelligence can produce inaccurate, incomplete, unexpected or outdated output.

Customers must not assume that an AI-generated response is correct merely because it appears confident or convincing.

AI output should be reviewed with a level of human oversight appropriate to the potential consequences of an error.

AgenticSA may not knowingly be used to:

• facilitate unlawful activity;
• commit fraud;
• impersonate another person unlawfully;
• deliberately deceive individuals;
• harass, threaten or intimidate;
• facilitate hate-based abuse;
• distribute unlawful spam;
• conduct phishing;
• distribute malware;
• obtain unauthorised access to systems;
• unlawfully collect personal information;
• circumvent consent or opt-out requirements;
• violate privacy rights;
• infringe intellectual-property rights;
• create unlawful discriminatory outcomes;
• facilitate exploitation or trafficking;
• generate unlawful sexual exploitation material;
• facilitate violence or serious physical harm;
• conceal material AI involvement where disclosure is legally required;
• manipulate vulnerable individuals through deceptive automated interactions; or
• otherwise violate applicable law.

Where an AgenticSA AI Employee is used for marketing, lead follow-up or promotional communications, the customer is responsible for complying with applicable direct-marketing requirements.

This includes, where applicable:

• establishing lawful authority to contact the individual;
• obtaining consent where required;
• maintaining appropriate consent records;
• identifying the business communicating with the person;
• providing required disclosures;
• respecting objections;
• respecting do-not-contact preferences; and
• providing and honouring appropriate opt-out mechanisms.

A customer may not use AgenticSA to bypass a person's withdrawal of consent or objection to direct marketing.

Customers using outbound AI Voice functionality are responsible for ensuring that each contact may lawfully be called.

Customers must implement appropriate procedures regarding:

• consent or other lawful authority;
• AI disclosure;
• caller identification;
• calling purpose;
• calling times;
• opt-outs;
• do-not-contact requests;
• frequency;
• call recording; and
• applicable industry requirements.

AgenticSA may impose additional technical or provider restrictions even where a customer believes a call is otherwise legally permissible.

Customers should avoid supplying unnecessary sensitive or special personal information to AI Employees.

Where processing of such information is necessary, the customer must determine that appropriate lawful grounds, security measures and safeguards exist.

AgenticSA should not be used to indiscriminately collect sensitive information simply because the technology is capable of doing so.

Customers must exercise additional caution before configuring AI workflows involving children.

Where the processing of children's information requires consent, authorisation or another lawful basis, the customer is responsible for establishing that requirement before using AgenticSA for that purpose.

AgenticSA may assist businesses operating in regulated industries, but an AI Employee should not independently replace a qualified professional where professional judgement, licensing or statutory responsibility is required.

Without appropriate qualified human review, AgenticSA must not be configured to provide definitive:

• legal advice;
• tax advice;
• regulated financial advice;
• investment advice;
• medical diagnosis;
• medical treatment advice; or
• other regulated professional advice.

AI may assist with administrative, informational, scheduling, intake, routing and approved support functions where appropriate.

Customers should not rely solely on AgenticSA to make significant decisions about individuals where an incorrect or unfair decision could materially affect that person's rights or interests.

Examples include decisions relating to:

• employment;
• credit;
• insurance;
• healthcare;
• legal rights;
• eligibility for essential services; or
• similarly significant matters.

Appropriate human review should be incorporated where required.

Customers are responsible for ensuring information supplied to train their AI Employee is:

• accurate;
• current;
• authorised;
• relevant; and
• appropriate for the intended use.

Customers should promptly update information that becomes inaccurate.

Customers must protect credentials, connected accounts and integrations.

Customers may not:

• attempt to bypass AgenticSA security controls;
• probe systems without authorisation;
• extract information belonging to another customer;
• intentionally compromise AI safeguards; or
• use AgenticSA to attack third-party systems.

AgenticSA and its technology providers may use authorised technical logs, conversation records and other operational information as reasonably necessary to:

• deliver the service;
• troubleshoot;
• improve configurations;
• investigate incidents;
• maintain security;
• provide support; and
• enforce applicable policies,

subject to applicable privacy and contractual requirements.

Creative Media Agency may restrict or suspend an AI Employee where reasonably necessary because of:

• suspected unlawful activity;
• serious abuse;
• security concerns;
• repeated policy violations;
• provider restrictions;
• excessive spam;
• fraud;
• material privacy risk; or
• a legal or regulatory requirement.

AgenticSA seeks to implement AI according to five practical principles:

Transparency — People should not be deliberately misled about material AI involvement.

Human Accountability — Businesses remain accountable for how AI is deployed.

Purpose Limitation — AI should be configured for legitimate, defined business purposes.

Data Minimisation — AI should process only information reasonably necessary for its intended function.

Human Escalation — AI should know when the appropriate next step is a person.

This Policy may be updated as AI technology, applicable laws and underlying provider requirements evolve.

Questions about responsible AI use may be directed to:

Creative Media Agency (Pty) Ltd

Email: hello@creativemediagency.co.za