Legal
How we use AI
Last updated 12 October 2026
Claims Suite uses generative AI. Here is exactly what it is used for, what it works from, what it cannot do, and where it can be wrong.
What we use it for
Claims Suite uses large language models to turn the notes, measurements, photographs, line items and documents you enter into first-draft written output — supplement requests, justification narratives, correspondence you may choose to send, and summaries of where a job stands.
What the AI is working from
It drafts from the information you supply and the work email you connect. Once your mailbox is connected, Claims Suite reads your work mail and its attachments, including carrier estimates, letters and adjuster emails sent to you, and files them on the right job. It also uses photos from accounts you connect (such as CompanyCam), plus reference material such as published building-code language, manufacturer installation specifications and standard estimating line-item descriptions. It does not have access to any insurance policy unless one is uploaded or emailed to you, and it has no access to carrier systems or to a carrier's internal claim files.
What the AI does not do
- It does not decide whether a claim is covered.
- It does not determine what a claim is worth.
- It does not communicate with any insurance carrier or adjuster.
- It does not send anything on its own. Mail goes out only when you press Send. With Gmail and Outlook the draft is saved to your Drafts folder and you send it from there. Mailboxes connected by IMAP cannot hold a draft for us, so when you press Send in Claims Suite it sends from your address for you.
- It does not make any decision about you, your customer, or any homeowner. There is no automated decision-making with legal or similarly significant effects.
Language models can produce inaccurate, incomplete, outdated or fabricated content — including incorrect code citations, incorrect measurements and incorrect prices. Every output is a starting point that requires your review. You are the author and the sender of every document, and you are responsible for verifying every factual statement, quantity, price and citation before you use it.
Human review is required, not optional
Claims Suite is deliberately built so that a contractor reviews, edits, approves and signs each document. Nothing is sent unless you press Send. Do not send any Claims Suite output unreviewed.
Your data and the model providers
Claims Suite uses Anthropic's Claude models through Anthropic's commercial API. Under Anthropic's published policy, inputs and outputs sent through its commercial API are not used to train its models by default, and Anthropic deletes API inputs and outputs from its systems within 30 days of receipt (see the Anthropic Privacy Center). We do not opt in to any Anthropic training programme. We do not sell your data and do not use it to train any model for anyone else's benefit.
Our Privacy Policy explains what we collect and how long we keep it. Business customers who need a data processing agreement can request one at support@claimssuite.com.
How Claims Suite learns
From your own history. When you connect your mailbox, Claims Suite reads the estimates, supplements and rebuttals you have sent, what carriers did in response, and how you write to adjusters, homeowners and suppliers. It keeps a private copy of that job mail (see the Privacy Policy) and learns which arguments, evidence and wording have worked for you with each carrier. It also compares what you actually sent with the drafts it wrote. Some lessons are held for your review before they are used. This learning is scoped to your account: your letters, prices, photos, customers and company name are never shown to another customer.
Shared, de-identified patterns. Claims Suite also counts which kinds of arguments carriers approve across many businesses — never who sent them or what the letter said — and uses those counts to make everyone's strategies stronger. A pattern is used only once it has been seen across at least three unrelated businesses, and data from Gmail accounts is never included. The account owner can turn this off in Settings → Shared learning. Details.
What we never do. We do not train AI models on your content, and we do not sell it.