AI Use Policy

This AI Use Policy explains how Operator uses artificial intelligence to run your business brain and the agents you hire. It is incorporated by reference into the Terms of Service and read alongside the Privacy Policy and the Acceptable Use Policy. It is not legal advice.

Operator runs autonomous agents that scan your connected systems, build a per-business knowledge store (your business brain), and draft and execute work on your behalf. AI does the heavy lifting:

• Reading and summarizing the data in the systems you connect.
• Drafting customer replies, review responses, posts, and reminders.
• Classifying signals and proposing the next action.
• Executing approved actions inside the OAuth scopes you grant.

Deterministic outcomes (billing, attribution, routing) are computed in code, not by a model. We use AI for judgment, not for math.

Agents act autonomously only within the limits you configure. You stay in control:

• Every action an agent takes is recorded in your audit log with the agent identity, the inputs, and the outcome.
• You can review, dispute, and roll back actions from your dashboard.
• You can pause an agent, narrow its scope, or revoke a connected account at the provider at any time.
• Actions that need human judgment (legal, regulated, or contractual decisions) are your responsibility to review before they take effect.

AI systems can be wrong. Agent output may contain errors, omissions, or text that reads confidently but is inaccurate. Operator is a business tool, not a substitute for professional legal, financial, medical, or tax advice. You are responsible for reviewing agent output before relying on it for any decision that carries legal, financial, or safety consequences. We do not warrant that agent output is accurate, complete, or fit for any particular purpose.

Your business brain is stored per business and isolated by row-level security keyed to your business identifier. With respect to AI:

• We do not use your private business data to train shared or third-party models.
• We do not sell your data, and customer brains never enter the public Substrate Data API.
• Where a model provider offers zero-retention or no-training terms, we use them for agent inference.

See the Privacy Policy for the full data-handling detail and the Security overview for isolation controls.

Agent inference runs on a short list of established model providers (currently Anthropic and OpenAI), plus self-hosted local models for bulk enrichment. Each provider is engaged as a sub-processor under data-protection terms no less protective than our own. The current list is on the Sub-processors page.

Some laws require you to disclose when content is generated by AI or when a customer is interacting with an automated system rather than a person. You are responsible for making any disclosure the law requires for your business and your jurisdiction. Configuring an agent to impersonate a human where disclosure is required, or to do anything else the Acceptable Use Policy prohibits, is a violation of that policy.

A verified win is an outcome an agent produced that we can attribute to that agent and confirm through a connected system of record (for example, a booked job traced to your calendar or a recovered invoice traced to a payment in Stripe). Wins are verified in code against real events, not self-reported by a model. The $1 per verified win fee applies only to wins that pass verification; failed attempts cost nothing.

We will update this policy as our use of AI evolves. Material changes will be communicated through the in-product notification surface or by email to your account address before they take effect.

Questions about how Operator uses AI: operator@operator.fyi. Operator AI LLC, Wyoming, USA.