AI and Cortex data use

Cortex is the AI built into Limnn. It answers questions about your workspace, summarises records, calls and conversations, drafts emails and replies, classifies and routes work, builds objects and workflows from a description, and runs AI agents that can propose or take actions. Cortex works only inside your workspace and only with data the signed-in user is allowed to see.

Each request carries only what that task needs. Cortex never sends data from one workspace into another workspace's request.

• Limnn does not use your workspace data to train or fine-tune AI models, and we do not use it to improve AI for other customers.
• Anthropic does not use inputs or outputs from its commercial API to train its models. Anthropic keeps API inputs and outputs for up to 30 days for trust and safety review. Limnn does not have a zero-data-retention agreement with Anthropic.
• We use Google's Gemini API as a paid service. Under Google's Gemini API Additional Terms of Service for paid services, Google does not use prompts or responses to improve its products.
• Google may keep Gemini API inputs and outputs for a limited period to detect abuse and meet legal obligations, under its own terms. Limnn does not have a zero-data-retention agreement with Google either.
• Results that Cortex produces (summaries, drafts, scores, generated content) are stored in your workspace like any other record, and you can edit or delete them.

When a user asks

Chat, “summarise”, “draft”, “generate”, Cortex Builder and similar actions run only when a user triggers them.

Automatically, when an admin turns it on

• Help-centre and website chat assistant: answers visitors from your knowledge base and escalates to a person; admins can switch AI off so chats go straight to agents.
• Case-hygiene rules: admin-configured rules that Cortex applies to tickets on a schedule (every few minutes).
• Flows and journeys that include an AI step.
• AI agents (playbooks) for support and sales.
• Beam recaps and assistance in channels, which admins can disable.
• Learning AI tutoring and course-authoring assistance, and recruiting CV parsing and screening summaries, where those features are used.

Every agent tool is classed by risk. Read tools run freely. Low-risk changes (for example adding a note or tag) run automatically only when an admin has set the agent's autonomy to allow it; in “draft” mode they are proposed instead. Risky actions: anything customer-facing or hard to undo, such as sending a message, always stop and wait for a person to approve them. Pending proposals appear in an approval queue, and every run, step and decision is recorded.

Cortex can produce scores and suggestions, for example lead scores, recruiting match scores, ticket priority, employee retention-risk indicators or productivity models. These are decision support for people in your organisation. Limnn does not make legal or similarly significant decisions about individuals on its own; customers that use these features are responsible for human review and for any notices their law requires.

AI use is metered in credits based on the provider cost of each request. Workspaces receive an included allowance (500 credits every 30 days on the free allowance, more on paid plans), can top up, and can set budgets per team. Agent runs reserve credits against a per-run spend cap; if credits run out, agents fall back to read-only or draft behaviour.

• Turn the help-centre chat assistant on or off.
• Turn Cortex in Beam on or off.
• Choose agent autonomy (draft, auto for low-risk, full) and which tools each agent may use.
• Enable or disable whole modules for the workspace.
• Set credit budgets per team.
• Restrict what each role can see with roles and field-level permissions. Cortex inherits those limits.

There is not yet a single switch that disables every AI feature at once; contact support@limnn.com if you need AI turned off for your workspace.

AI output can be wrong or incomplete. Review it before relying on it or sending it to anyone. Ownership of inputs and outputs is covered in the Terms of Service, and acceptable use of AI in the Acceptable Use Policy.