Last updated: September 11, 2026
Frontminder helps businesses build AI chat agents — software that generates replies automatically using large-language models. This page explains, in plain terms, how those agents work, what they can and cannot do, and the responsibilities that come with operating one. It supplements our Terms of Service and Privacy Policy.
How the agents work
An agent answers using a combination of the instructions and knowledge you give it and a third-party language model that generates the wording of each reply. Optionally, an agent can search your knowledge base for relevant content, call tools you enable (such as capturing a lead), and route to a person. Model providers are listed on our Sub-processors page.
What AI output is — and is not
- It is generated, not retrieved verbatim. Replies are produced word by word and can be inaccurate, outdated, incomplete, or occasionally fabricated (“hallucinated”), even when the underlying knowledge is correct.
- It is not professional advice. Agents must not be presented as a lawyer, doctor, financial adviser, or other licensed professional, and their output should not be relied on as such.
- It is your responsibility. You configure the agent, supply its knowledge, and choose where it runs. You are responsible for reviewing its behavior and for the messages it sends on your behalf.
Transparency to end users
You should not deceive people into thinking they are talking to a human when they are talking to an AI agent. We recommend a clear disclosure — for example, a short note that an assistant is automated and that a person is available — and, in jurisdictions or channels that require it, such disclosure may be mandatory. Do not configure an agent to deny that it is an AI when asked directly.
Human oversight
Keep a path to a human for situations the agent should not handle alone. Do not use agents to make automated decisions that produce legal or similarly significant effects on a person (for example eligibility, credit, or employment decisions) without appropriate human review and any disclosures the law requires.
Consent and messaging rules
If you connect an agent to a messaging channel or use it for outbound or marketing communication, you are responsible for complying with the rules that apply to that channel and to your customers’ locations. This includes obtaining any required consent before messaging, honoring opt-out and unsubscribe requests promptly, identifying the sender, and respecting quiet-hours and frequency rules where they apply. You must have the necessary rights and consents for any contact data you upload.
Prohibited uses
You must not use Frontminder agents to: impersonate a real person or organization without authorization; generate or distribute unlawful, harassing, hateful, sexual-exploitation, or infringing content; facilitate fraud, phishing, or malware; provide instructions for serious physical harm; scrape or harvest personal data unlawfully; or evade a platform’s or a person’s stated preferences. These are in addition to the acceptable-use rules in our Terms of Service.
Accuracy and safety practices we recommend
- Give the agent accurate, current knowledge and review it regularly.
- Test the agent’s answers to your most important questions before going live.
- Use the optional second-pass validation and personal-data masking features where appropriate to your use case.
- Monitor conversations and correct instructions when the agent gets something wrong.
Data used for AI
Message content is sent to a model provider to generate a reply and may be optionally masked to remove certain personal data beforehand. We do not use your end-user conversation content to train our own foundation models. See the Privacy Policy and Sub-processors page for details on how data flows.
Changes
We may update this page as the Service, the models, and the applicable rules evolve. Material changes will be signaled by updating the date above.
Contact
Questions about responsible use of the Service can be sent to support@frontminder.com.