AI use disclosure

1. The short version

• AI drafts; an attorney approves. Nothing the AI produces is sent outside the firm until a person at the firm approves the exact content.
• No legal advice to the public. InjuryMcGee is a tool for law firms. It does not advise injured people or anyone else, and no AI feature is offered directly to the public.
• Models run through Microsoft Azure, in Azure resources we control.
• No training on your data. Not by us and not by our model provider.

2. What the AI does

• Classifies incoming documents and suggests which case they belong to.
• Extracts facts such as dates of service, providers, amounts, claim numbers and deadlines-triggering events.
• Summarizes medical records into a chronology with page citations.
• Drafts letters, records requests, demand letters and discovery shells from the case file and the firm's templates.
• Answers staff questions about a case through the McGee assistant, with citations to the source.
• Helps with first-pass research, listing authorities only after their citations are checked against a public case-law database.

Some work involves no AI at all. Deadline arithmetic, ledger math, the approval gate and the checks that block sending to the wrong recipient are ordinary deterministic code.

3. What the AI does not do

• It does not send email, letters, faxes or texts. Only the approval gate can release an outbound item, and only as approved.
• It does not file anything with a court.
• It does not negotiate with insurers or opposing counsel, decide what a case is worth, or accept or reject offers.
• It does not create an attorney-client relationship with anyone or give legal advice.
• It does not set deadlines on its own. A deadline needs a rule citation or a quoted source line, and a person confirms it.

4. Which models, and where

InjuryMcGee uses large language models, including Anthropic's Claude models, accessed through Microsoft Azure AI Foundry, and Azure AI Document Intelligence for OCR. These run in Azure resources owned by Felican AI Inc. We pick the least powerful model that does each job well: small models for sorting and extraction, stronger models for drafting, the strongest for citation checks. The specific models may change as better or cheaper ones become available; this page will be updated when the provider changes.

5. What the models see

• Only the part of a case a task needs, such as a single document or the ledger lines for a demand.
• Never another firm's data, and never another client's file for a different case.
• Social Security numbers and bank or trust account details are not sent to models.
• By default we log the size and cost of each model call, not the prompt or response text.
• Incoming documents are treated as untrusted data; text inside a document cannot instruct the system to send anything.

6. Accuracy and limits

AI can be wrong, including confidently wrong, and can misread poor scans or handwriting. InjuryMcGee is designed to show its sources and to say “I don't know” rather than guess, but those safeguards reduce errors; they do not eliminate them. Every AI output must be reviewed by a qualified person before it is relied on. See See section 6 of our Terms of Service.ldquo;AI outputSee section 6 of our Terms of Service.rdquo; in our Terms of Service.

7. Controls you have

• Turn individual AI features on or off for your firm.
• Choose which roles may approve which kinds of outbound items.
• Set a monthly AI spending ceiling.
• Review AI usage and cost per case in Reports.

8. Professional rules

Bar guidance on generative AI, including The Florida Bar's Ethics Opinion 24-1, emphasizes confidentiality, oversight of the work, reasonable fees and transparency with clients. InjuryMcGee is built to support those duties with approval gates, audit logs and usage reporting, but compliance remains the firm's responsibility. Firms decide how to disclose AI use to their clients. Any client-facing message drafted with AI that a firm chooses to send is labeled as AI-assisted where the firm's settings require it.

9. Contact

Questions about how we use AI: [email protected].