AI and automation
Last updated 4 November 2026
Brand Autopilot uses AI to save you time, and people to make the decisions. This page explains where AI is used, what it is given, and what it may and may not do on its own.
The model we use
We use Claude, made by Anthropic. If the main model is busy, we may try once with another Claude model. Under Anthropic's commercial terms, what we send is not used to train its models. Each plan has a monthly AI budget, so AI use stops at a set limit rather than growing without bounds.
Where AI is used
| Task | What the AI is given | Who decides |
|---|---|---|
| Post drafts and repurposed versions | The brief, the brand blueprint and the target network's rules | A person reviews and approves every post |
| Weekly content plan | The brand's pillars, goals and recent plan | A person keeps, edits or drops each idea |
| Brand Blueprint drafts | Interview answers, and the website text, documents or past posts you chose to share | A person edits and saves the blueprint |
| Company profile at sign-up | The company's name, website text and social handles; the AI may run a few web searches | The owner checks it before it is saved |
| Suggested comment replies | The comment, the post it is on and the brand blueprint | A person, unless the brand uses automatic replies |
| Weekly growth brief | The company's own publishing and engagement numbers | Suggestions only; nothing is changed |
Posts are never published by AI alone
Every post, broadcast and article needs a person's approval before it is published, and that approval is tied to the exact text reviewed. Drafts are checked against the brand blueprint and for phrases that sound machine-written, but they can still contain mistakes, so please read them.
Comment replies: three modes
Each brand chooses how its inbox answers comments on Facebook, Instagram, LinkedIn and YouTube:
- Off (the default): people write every reply.
- Draft for review: the AI suggests a reply to each new comment, and a person approves, edits or discards it.
- Reply automatically: the AI replies on the brand's behalf, within a daily limit (30 replies in 24 hours by default, set by the company). Anything the AI or our checks consider risky, such as complaints, legal or sensitive topics, or a reply that breaks the blueprint, is held back for a person.
Automatic replies are recorded as automatic in the company's audit log. The company can switch a brand back to Off at any time.
Untrusted text
Comments, web pages and documents are written by other people. We pass them to the AI only as quoted material, not as instructions. The AI has no way to change settings or reach another company's data, and posts still need a person's approval. AI models can still be misled, which is why automatic replies hold risky comments back for a person and run within a daily limit.
Your responsibility
You remain responsible for what is published under your brand, including automatic replies you turned on. Where a network or the law requires AI-generated or promotional content to be labelled, you must label it. See our terms of service and acceptable use policy.
Questions
See how we handle data in our privacy policy, or contact us.