AI & Automation Policy

AI is optional and purpose-specific

Not every BNAX project uses AI. We first identify the task, intended benefit, affected people, information involved, and an appropriate review process. Straightforward organization or conventional automation may be a better fit.

The current website inquiry and proposal workflow does not automatically send your answers to an AI model. Its proposal suggestions are service templates reviewed by the owner. This policy explains expectations for separately agreed AI-assisted services, not a hidden AI feature in the inquiry form.

Scope, providers, and information

Before an AI workflow is commissioned, its scope should identify the provider, authorized data, access permissions, human reviewer, costs, usage limits, retention settings, and relevant provider terms. A product name alone is not a commitment about training, confidentiality, data residency, or retention.

Do not put confidential, personal, student, medical, payment, or regulated information into an AI tool merely because it is available. Confirm authority, necessity, contractual protections, provider settings, and an appropriate transfer method first. Use de-identified or synthetic examples when practical.

Any promise that a provider will not train on particular data must be supported by the applicable account configuration and contractual terms. BNAX does not make a blanket promise about every external AI product.

Human review and limitations

AI output can be inaccurate, incomplete, biased, outdated, or similar to existing material. Review factual claims, calculations, sources, permissions, accessibility, and suitability before relying on or publishing it. AI suggestions are not a substitute for professional judgment or an authorized decision maker.

Financial, legal, medical, employment, educational, and other consequential decisions require a suitable human decision process and any legally required safeguards. We do not present a draft, model score, or automated recommendation as a verified professional determination.

Automated communications, purchases, deletions, account changes, and other external actions need explicit scope, authorization, and appropriate controls. A workflow should define what needs approval and what happens when a step fails or the output is uncertain.

Testing, monitoring, and stopping

Agree representative test cases and acceptance criteria before production use. Tests should address incorrect inputs, service outages, duplicates, unexpected output, permissions, and recovery where relevant. A working demonstration does not prove safe operation in every real situation.

The delivery plan should name who monitors the workflow, handles alerts, reviews output, and can pause it. Provider or model changes may require renewed testing. Ongoing monitoring is included only to the extent specified in the support agreement.

If a workflow produces an unexpected or harmful result, stop relying on that output, use the agreed fallback, and report the issue through the project support channel. Do not continue a consequential automated action merely to see whether the next result improves.

Content rights, disclosure, and training

Clients must have authority to supply source materials. The agreement should distinguish original human work, AI-assisted material, third-party assets, and software licenses. Exclusive rights or copyright protection for AI output are not guaranteed by this policy.

Use disclosure when required by law, platform rules, an audience’s context, or the agreed project standard. Do not use AI to fabricate credentials, testimonials, research, or a person’s endorsement. Training examples are educational and must be adapted and reviewed before operational use.

Bring AI-related questions or restrictions to contact@bnaxdesigns.com before an engagement so they can be included in the written scope.