AI & data use

Last updated: November 2026

Draft — to be reviewed by counsel before publication. This document is not legal advice and may change.

This document explains in plain terms how AI works inside Squadre, which data passes through it, what it may and may not do on its own, and what always stays in your hands.

1. What the AI does

Squadre's agents read the conversations and events you connect, write a hint line, draft replies, open and assign tasks, and produce insights. Each agent has a defined domain and a written guardrail — for example, the Revenue agent never changes prices.

2. A person approves, by default

Every type of action starts at "a person answers": the agent drafts and you send. Automatic sending switches on only when the account owner explicitly moves a type of action up the trust ladder, and it can be moved back at any time.

In a conversation flagged as an escalation (an upset customer, a refund demand) the agent stops replying and hands over to a person.

3. What is sent to the model

To draft a reply we send the model provider the relevant conversation excerpts, the booking or task details that matter, and the business profile and playbook you defined. We send only what the specific action needs.

Model provider: Anthropic (USA), through its commercial API
Channel tokens, passwords and payment details are never sent to the model

4. We don't train on your data

Squadre does not use your conversations, contacts or content to train models. Under the model provider's commercial terms, it does not train on data sent through the API either.

One business's data is never used to answer another business's customers.

5. Mistakes and responsibility

Language models can be wrong or imprecise. That is why the default is a draft for approval. You set the level of autonomy and are responsible for content sent in your business's name.

The AI does not make solely-automated decisions with legal or similarly significant effects on a person — no credit scoring, no candidate screening, no personalised pricing.

6. Transparency toward your customers

You are responsible for telling your customers about the use of automated tools where the law that applies to you requires it. We recommend a short line in your own privacy notice and are happy to help word it.

7. Usage cap and oversight

Each business has a monthly AI usage cap. When it is reached the agents pause and nothing is billed beyond it. Every agent action is recorded in a log you can review.

8. Cross-border transfers

Model processing and system hosting take place in the United States. Transfers follow the Israeli Privacy Protection Law and its transfer regulations, and are subject to our sub-processors' contractual commitments. Details are in the DPA.

9. Your controls

You can switch off an agent, switch off AI altogether and keep using the unified inbox, export your data and request deletion — at any time.