AI Output Accuracy, Human Review & Optimisation Notice#
AI WorkerX Sdn. Bhd.
Company No.: 202501054192 (1655598-U)
Last Updated: 15 September 2026
Effective For All Users Worldwide
1. Purpose and scope#
This Notice explains the limitations of artificial intelligence and how AI WorkerX Sdn. Bhd. (“AI WorkerX”) and its customers should manage AI-generated outputs, automated responses, suggested actions and post-deployment optimisation. It applies to AI WorkerX subscriptions, add-ons, response top-ups, prototypes, pilots and production solutions that use artificial intelligence or automated decision support.
“Output” includes generated text, summaries, classifications, recommendations, extracted information, draft documents, customer replies, tasks, routing decisions, workflow suggestions and other AI-assisted results.
2. AI-generated outputs may contain mistakes#
AI systems generate outputs by identifying patterns and probabilities. They do not possess perfect knowledge or guaranteed understanding. An output may therefore be inaccurate, incomplete, outdated, inconsistent, misleading, unsuitable for a particular situation or different from the customer’s intended result.
An output may also sound confident even when it is incorrect. Connecting an AI system to a knowledge base, search tool, database, API or workflow can improve usefulness and verification, but does not guarantee that every output or action will be correct.
3. Appropriate human review and monitoring#
Customers must apply a level of human review, approval and monitoring that is appropriate to the importance, risk and potential impact of the use case. Important outputs should be checked against reliable source material before they are relied on, shared externally or used to take action.
Where a customer enables automatic replies, automatic task creation, automated routing or other actions without prior human approval, the customer should select suitable low-risk use cases, configure escalation rules, monitor performance and maintain a process for correcting inappropriate outputs or actions.
4. Professional, regulated and high-impact matters#
AI WorkerX is intended to support productivity and operations. Unless expressly agreed in a separate written professional engagement, outputs are not legal, financial, accounting, tax, audit, medical, investment or regulatory advice and do not replace the judgement of a suitably qualified professional.
AI outputs should not be used as the sole basis for a decision that may materially affect a person’s legal rights, employment, credit, insurance, healthcare, education, housing, financial position or access to essential services.
5. Real-world go-live and optimisation#
Testing and prototypes cannot reproduce every real-world conversation, edge case, customer expression, language variation, data-quality issue, integration behaviour or staff practice. During the first weeks after go-live, actual use may therefore identify areas requiring configuration, knowledge-base, prompt, workflow, escalation or integration adjustments.
The discovery of an improvement opportunity does not automatically mean that the entire system is defective. AI WorkerX and the customer should review the specific example, expected outcome, source data, system setting and operational context before deciding the appropriate response.
Any optimisation period, included support work, service level, response time and implementation commitment will be governed by the applicable quotation, order form, statement of work, subscription plan or support arrangement.
6. How issues and improvement requests are handled#
| Issue category | Typical meaning | How it is handled |
|---|---|---|
| Reproducible defect | The solution materially fails to perform an agreed documented function under the agreed conditions. | Investigated and addressed under the applicable agreement, support plan or warranty commitment. |
| Tuning or optimisation | Adjustment to prompts, rules, knowledge content, tone, thresholds, routing, exceptions or workflow configuration within the agreed scope. | Prioritised and implemented according to the agreed onboarding, pilot or support arrangements. |
| Change request | A new feature, integration, workflow, data source, channel, module, report or material change to the agreed requirements. | May require separate scoping, timeline, quotation and written approval. |
| Third-party or customer dependency | An issue caused by inaccurate or incomplete customer data, customer instructions, messaging channels, model providers, APIs, payment providers, cloud services or other external systems. | The parties cooperate to investigate; additional work or third-party remediation may be required. |
7. Customer responsibilities#
- Provide accurate, lawful and sufficiently complete instructions, data, policies, source material and access required for the agreed use case.
- Review and approve knowledge-base content, workflows, business rules, escalation paths and permitted automation settings.
- Apply appropriate human oversight and professional review for important, regulated or high-impact matters.
- Monitor live use and report material issues promptly with concrete examples and the expected result.
- Ensure staff and authorised users understand the system’s role, limitations and escalation procedures.
- Maintain reasonable backups, business continuity procedures and alternative processes for critical operations.
8. AI WorkerX responsibilities#
- Use commercially reasonable care in configuring and supporting the solution in accordance with the applicable agreement and agreed scope.
- Investigate reproducible issues reported with sufficient information and distinguish defects, tuning requests, change requests and third-party dependencies.
- Implement agreed fixes or optimisation work in accordance with the applicable support, onboarding or project arrangements.
- Maintain reasonable security, access controls and data handling practices as described in the applicable terms and privacy documentation.
- Communicate material known limitations that are relevant to the agreed use case when reasonably practicable.
9. No guarantee of perfect accuracy or a particular outcome#
AI WorkerX does not promise that every AI-generated output will be accurate, complete, unique, uninterrupted, error-free or suitable for every purpose. AI WorkerX also does not guarantee revenue growth, cost savings, regulatory compliance, legal success, professional outcomes, customer conversion, investment results or any other particular business outcome.
This Notice is not intended to excuse fraud, wilful misconduct, breach of an express written commitment, failure to address an agreed reproducible defect, or any liability that cannot lawfully be excluded or limited.
10. How to report an issue#
Report issues through the support channel identified in the customer account, quotation, order form, statement of work or support plan. To help AI WorkerX investigate, provide where available:
- the date and time of the event;
- the relevant account, conversation, transaction, task or job identifier;
- the input, source data or instruction provided;
- the output or action produced;
- the expected output or action;
- screenshots, logs or supporting source material; and
- the operational impact and urgency.
11. Relationship with other terms#
This Notice forms part of the AI WorkerX Terms & Conditions when it is incorporated by reference at checkout, in an order form or in another agreement. If there is a conflict, the signed order form or statement of work, followed by the applicable Terms & Conditions, will prevail to the extent of the conflict.
Nothing in this Notice excludes or limits any right, remedy, guarantee or liability that cannot lawfully be excluded or limited. AI WorkerX may update this Notice prospectively. The version accepted by the customer should be retained with the corresponding transaction record.