SimpleOS Desk AI Use Statement

Last updated 13 November 2026

This statement says how SimpleOS Desk uses AI models: what is sent to a model, where the model runs, how the result is marked, and what a person must do before the result counts. It is part of our Terms of Service. "We" means Simple Intelligence Group, Inc.

1. The short version

AI in SimpleOS Desk writes drafts and suggestions. A person reviews them before they take effect outside your workspace.

Every record an AI feature writes is marked as AI-generated and carries a confidence level and a reason.

Models run inside Microsoft Azure, through Azure AI Foundry. Your content is not sent to a model supplier's own public service.

We do not use your data to train any AI model.

2. Where the model runs

SimpleOS Desk uses two Microsoft Azure services. The help assistant runs on Azure OpenAI, with a model supplied by OpenAI. Agentic resolution and predictive alerts, in workspaces where they are enabled, run on Azure AI Foundry, with Claude models supplied by Anthropic.

The model is reached as a Microsoft Azure service. Microsoft's terms for that service apply to how it handles a prompt while producing a response.

3. What is sent to a model

A feature sends content to a model only when someone uses that feature, or when an administrator has switched on an automation that uses it. A feature that is not enabled for your workspace sends nothing.

What SimpleOS Desk sends: The question a user types into the help assistant. In workspaces where agentic resolution or predictive alerts are enabled, the ticket and device details those features work on. While a feature is not enabled for a workspace, no prompt is sent for it.

The prompt contains the content needed for the task and the instructions that tell the model what to do with it.

4. What is not sent to a model

Passwords and sign-in credentials are never sent. We do not hold your Microsoft password at all.

Access tokens and keys for the services you connect are never sent.

Payment card numbers are never sent. We do not hold them.

Content from another customer's workspace is never included in a prompt for yours.

5. How AI output is marked

A record that an AI feature writes is stored with three facts: that it was AI-generated, how confident the model was, and a short reason for what it produced. The product shows these with the record so that a reviewer can judge it.

6. What we keep

The model service processes a prompt to return a response.

Inside the product, the record an AI feature writes is part of your workspace data, together with its confidence level and reason. Where the product keeps the prompt or the response so that a person can trace how a record came about, that copy is also your workspace data. It is handled on the same terms as the rest, including export and deletion.

7. A person reviews first

AI output is a proposal. Before it has an effect outside your workspace, a person reviews it. Examples of an outside effect are a message sent to someone and a change made in a system you have connected.

Where the product asks for approval before such a step, the step does not happen until a person approves it.

Decisions about who may see or change a record are never made by a model. They are made by fixed rules in the product, on the server.

8. AI agents and your data

Where an AI agent in the product reads or changes workspace data, it does so only through the product's own controlled interface. That interface applies your workspace identifier itself, so an agent cannot reach another workspace by asking for it. Every call an agent makes is recorded with the workspace, the caller and the function used, whether or not the call succeeds.

9. Your part

Review AI output before you rely on it or send it to anyone outside your organization. It can be wrong or incomplete.

Do not treat AI output as legal, financial, medical or compliance advice.

Do not use AI output to make a decision that has a legal or similarly serious effect on a person unless a human has reviewed it.

Do not arrange for an approval step to be passed without a person reading what is being approved.

10. Changes

If a feature starts to send a new kind of content to a model, or if the model service or supplier changes, we update this statement and its date. If the change adds a company that handles your data, we tell workspace administrators by email at least 30 days before it takes effect.

11. Questions

Write to support@simpleintelligencegroup.com. For a privacy question, write to privacy@simpleintelligencegroup.com.

See also: Terms of Service, Privacy Policy, Data Processing Addendum, Trust.