NUMEXS / LEGAL / 08
AI & Automation Disclaimer
Practical limitations, review points, and responsibilities when AI tools or automated workflows are implemented.
What this page addresses
NUMEXS Corp offers AI Tools Integration and Workflow Automation among its services. A project may connect an approved AI product to a business task or configure a rule-based sequence between existing applications. The tool, inputs, outputs, and degree of autonomy must be defined for the specific use case.
AI integration is a technical service, not a claim that an AI model will be correct, unbiased, available, or suitable for every decision. Automation can also fail when a trigger, input format, permission, API, or vendor feature changes. The design should make these dependencies visible to the client.
Accuracy and human judgment
Generated text, classifications, summaries, recommendations, and extracted data can be incomplete, inaccurate, outdated, or misleading. An output should be checked by a qualified person before it is used in a consequential business process, sent externally, or entered into a system of record. The level of review should match the risk of the use case.
The client remains responsible for its business decisions, professional obligations, and final use of AI-assisted material. A technical implementation does not provide legal, medical, accounting, financial, or other regulated professional advice. NUMEXS should not represent a model response as a verified fact without an appropriate review process.
Data and confidentiality
The client should decide what information an AI or automation tool may receive. Sensitive personal information, confidential business data, credentials, and protected customer records may require restrictions, redaction, or a different method. Provider policies can govern retention, training, geographic processing, and administrator access.
NUMEXS should use only the data and vendor accounts approved for the project. The client should review a third-party AI provider's terms before allowing business data into that provider's service. A public promise about all model providers would be inaccurate because their terms and settings differ and can change.
Scope and testing
An accepted project scope should define the intended task, input sources, trigger conditions, output destination, exceptions, and human approval point. Representative test cases should include ordinary use and known edge cases. Test results show how a configured workflow behaved under those cases; they do not prove that every future input will be handled correctly.
A client should designate an owner for monitoring operational outcomes and reporting errors after handoff. Vendor models, APIs, pricing, or policies may change and require a later adjustment. Ongoing monitoring or retraining is part of a service only if explicitly included in the agreed support scope.
Automated actions and safeguards
An automated workflow can create records, send messages, update fields, or invoke an external service. The parties should identify which actions are permitted, what happens when an input is missing, and when an action requires human approval. The scope should also say who can pause or change the workflow.
A rollback path, logs, alerting, or manual alternative may be appropriate depending on the environment and risk, but these features are not automatically included in every service. They should be defined and priced in the accepted scope. The client should keep its own records required for compliance or business continuity.
Third-party platform limits
NUMEXS may configure a third-party AI or automation platform but does not control its model weights, global service availability, pricing, content rules, API rate limits, or independent data practices. The platform may produce different outputs for similar inputs or change behavior after an update.
The client is responsible for maintaining subscriptions, licenses, and rights to submitted information. External service charges are separate from the listed NUMEXS implementation fee unless a written proposal says otherwise. A platform-specific guarantee must come from the relevant vendor or a separately agreed contract.
Prohibited or high-risk uses
The suitability of a proposed AI use depends on law, sector, location, data type, and the consequences of an error. Before use in employment, credit, health, safety, legal rights, or another high-impact decision, the client should obtain qualified review of applicable obligations and controls.
NUMEXS should not be understood to approve a regulated or high-risk use merely by implementing an ordinary technical connection. If the client proposes such a use, the parties should define the required professional review, approvals, restrictions, and contractual scope before proceeding.
Updates and questions
This disclaimer describes the risks relevant to the services listed on the site at the update date. It may be revised as products, law, or project practices change. A signed scope and the third-party provider's terms govern the technical work and platform use for a particular engagement.
Questions about an AI or automation use case can be sent to the contact details below. Describe the data involved, the decisions affected, and who will review the output, without including sensitive records in a first enquiry.