AI and BYO-AI policy
Anvil Field works with zero AI. Capture, proof, memory, and every deterministic calculator run without a model. AI is a fenced drafting and translation convenience plus a clean export path to your own AI account.
Drafting, summarizing, translating the user's own text, explicit and rate-limited, with a check-before-send step.
A structured export of the account's own data that the user can take to their own ChatGPT/Gemini/Claude account.
You decide; Anvil Field remembers, computes, and proves.
Bound Camera, Offline Save Vault, and local-save trust never depend on a model.
Every built-in calculator is deterministic arithmetic over built-in tables. AI may summarize around accepted outputs; it never produces or modifies a computed value.
Nothing goes to any model without an explicit tap on a clearly labeled action.
Every output is a draft behind a check-before-send step. AI text never auto-saves into a record and never auto-sends to a customer.
No generated safety advice, hazard text, PPE guidance, or compliance statements.
Each action is a labeled button, and first use says text is being sent to Anvil Field's AI service.
Daily AI drafts used - write it yourself or use the BYO-AI export. Capture and calculators are never metered.
AI output renders in an editor with Draft — read and edit before you use it.
Photos, customer names, addresses, and gate codes are excluded from AI payloads by default. Gate codes are excluded always.
No AI statement about compliance, code, safety, legality, or correctness of work.
No AI-generated message leaves the account without the user reading and tapping send.
Customer photos, notes, addresses, voice, and account content are never used to train or fine-tune any model.
No account-wide scanning, no always-on assistant, no background enrichment of records.
AI cannot create, edit, or delete a record; it can only propose text the user explicitly accepts.
A structured backup file (or ZIP for larger scopes) of the account's own data at the scope the user picks.
Text and structure. Photos are referenced by filename/label by default; image files require an explicit ZIP option.
Generated as a local file, unlimited, free at every tier.
A live integration. Anvil Field does not call OpenAI/Google/Anthropic with your data and holds no API relationship on your behalf at launch.
The data goes to that provider under that provider's terms, in the user's own account. Anvil Field is not a party to it.
Before the file is written, a contents summary lists what's inside, and a required checkbox confirms the user understands provider terms.
This export is your data, for your own AI account. Check it for customer details and access codes before you paste or upload it — once it leaves Anvil Field, the AI provider's terms apply.
Pasting third-party AI text back into Anvil Field makes it a user-authored note like any other.
Anvil Field's AI can draft a summary, shorten your notes, or translate between English and Spanish - always from text you wrote, always as a draft you read and edit before anyone sees it. It runs only when you tap an AI button, and drafts are limited per day on each plan.
It will never: send anything for you, change your records, judge whether work is safe or up to code, or train on your data. Your photos, customer names, addresses, and gate codes stay out of AI requests.
Prefer your own AI? Export your data anytime and use it in your ChatGPT, Gemini, or Claude account. That file is yours - check it for customer details before you share it, because once it leaves Anvil Field, the AI provider's terms apply.
KEEP.
KEEP (preferred path).
DEFER, explicit per-use opt-in only.
DEFER.
KILL.
Anvil Field's AI is two honest things: fenced drafting/translation convenience inside the app, and a clean structured exit ramp to the user's own AI.