AI Statement
Effective date: September 24, 2026
Kestrel uses artificial intelligence. Here's exactly where, and what that means for you.
Where AI is used
Reading photographs of printed lists. When you scan a VIP list, lead sheet, or guest form, the photograph is sent to an AI model (currently Claude, made by Anthropic) which reads the printed text and returns names, phone numbers, vehicles, appointment dates and times, and similar fields as data. That's the only thing AI does in Kestrel.
Where AI is not used
- AI does not write, choose, or send your messages. Templates are written by you (or shipped as defaults you can edit), filled in with the customer's name and appointment details by simple text substitution, and sent by you.
- AI does not decide whom to contact, when, or how often.
- AI does not make calls.
- AI is not used to analyze your behavior, score customers, or make any automated decision about a person.
Accuracy
AI reading of printed text is good but not perfect. Digits can be swapped, names misspelled, rows merged or split. Kestrel reduces this by sending high-resolution images, flagging readings the model reports as uncertain, and letting you view the original photo next to the extracted number. Always check before you rely on a number or a name, especially where the app shows a "Check number" flag.
What the AI provider receives
The photographs you scan, and a written instruction describing the table layout. Photographs contain whatever is printed on the page — typically customer names, phone numbers, and vehicle details. Under Anthropic's API terms, this data is not used to train Anthropic's models. Photographs are not stored by Aerie Labs on our servers; the extracted text is.
Your responsibility
Because AI output can be wrong, you're responsible for reviewing extracted data before contacting anyone. Aerie Labs doesn't guarantee the accuracy of AI readings (see Terms of Service, sections 6 and 10).
How we disclose this in the app
- Before your first scan, the app tells you that photos are read by AI and that readings should be checked.
- Every uncertain reading is flagged in the list ("Check number") with a one-tap view of the original photo.
- Marketing claims about scanning speed or accuracy are meant literally and are subject to the limits described here; we don't claim AI reads perfectly, and we don't use AI to imitate a person or to generate messages that appear to come from someone else.
These practices follow the Federal Trade Commission's guidance on truthful, non-deceptive use of AI.
Changes
We may change AI providers or models to improve accuracy or cost. We'll update this statement when we do.
Questions: support@aerie-labs.com